Methods of predicting properties of differentiated neuronal cells

Abstract

Provided herein are methods for predicting properties, such as activity, function, and / or differentiation state, of an in vitro population of neuronal cells (e.g., a population of neuronal progenitor cells), as well as methods for selecting and / or implanting an in vitro population of neuronal progenitor cells having a desired property. Also provided herein are computing devices for performing the provided methods, as well as related compositions, articles of manufacture, and kits, including uses in methods of treating a subject having a neurodegenerative disease (e.g., Parkinson’s disease).

Description

Cross-reference to related applications

[0001] This application claims priority to U.S. Provisional Application No. 63 / 598,533, filed November 13, 2023, entitled “METHODS OF PREDICTING CHARACTERISTICS OF DIFFERENTIATED NEURONAL CELLS AND RELATED COMPOSITIONS OF DIFFERENTIATED CELLS”, the contents of which are incorporated herein by reference in their entirety for all purposes. Technical Field

[0002] This disclosure relates to methods for predicting the properties (such as activity, function, and / or differentiation state) of in vitro neuronal cell populations (e.g., neuronal progenitor cell populations), and methods for selecting and / or implanting in vitro neuronal cell populations with desired properties. This document also provides computing devices for performing the provided methods, as well as compositions, articles, and kits for use in methods of treating subjects with neurodegenerative diseases (e.g., Parkinson's disease). Background Technology

[0003] Parkinson's disease (PD) causes frailty-motor complications, and there is currently no available restorative treatment. PD is the second most common neurodegenerative disease after Alzheimer's, affecting approximately 0.3% of the general population and 1-2% of those aged 65 and older. With the aging populations in developed countries, the prevalence of PD is projected to double or triple. (Cha et al., (2023)) J. Mov.Disord .16: 22-41; Rong et al., (2021) Neurology 97: e1986–e1993; Dorsey and Bloem (2018) JAMA Neurol .75:9-10; de Lau and Breteler (2006) Lancet Neurol .5: 525-535).

[0004] Dopamine deficiency, resulting from the progressive loss of dopaminergic neurons in the substantia nigra, is a common feature of Parkinson's disease (PD). By the time of diagnosis, patients have already experienced significant nigrostriatal degeneration. Currently available treatments, such as dopamine replacement therapy (e.g., using levodopa or dopamine agonists), are beneficial to some patients, but have a limited therapeutic window due to side effects and reduced efficacy. Cha et al., Weiss et al. (1971) Lancet .1:1016-1017; Kang and Fahn (1988) Ration.Drug Ther .22: 1-7.

[0005] Cellular replacement therapies aimed at restoring lost dopamine-producing neurons have been developed for many years. A challenge in developing cell-based therapies for Parkinson's disease (PD) has been identifying suitable cell sources for neuronal replacement. One approach is the transplantation of fetal midbrain dopamine neurons, such as those performed in over 300 patients worldwide. Brundin et al., (2010) Prog. Brain Res .184:265-94; Lindvall and Kokaia (2010) J. Clin.Invest 120:29-40. Therapies using human fetal tissue in these patients have demonstrated evidence of DA neuron survival and in vivo DA release up to 10 or 20 years after transplantation in some patients. However, in many patients, fetal tissue transplantation cannot replace DA neuron function. Furthermore, Parkinson's disease patients treated with fetal cell transplantation sometimes experience transplant-induced motor dysfunction upon discontinuation of treatment. Evidence suggests that this serious side effect is caused by serotonin (5-HT) produced by the transplanted fetal cells. Tabar et al., (2011) Mov. Disord .26: 1997–2003. Furthermore, fetal tissue transplantation is plagued by challenges, including low quantity and quality of donor tissue, ethical and practical issues regarding tissue sourcing, and an unclear definition of transplanted cell heterogeneity, all of which contribute to variable clinical outcomes. Mendez et al., (2008) Nature Med 14: 507-509; Kordower et al., (1995) N. Engl. J. Med 332:1118-24; and Piccini et al., (1999) Nature Neuroscience 2:1137-40. Hypotheses regarding the limited efficacy observed in human fetal transplantation trials include: fetal transplantation may not provide a sufficient number of cells at the correct developmental stage, and fetal tissue is difficult to define by cell type and variables regarding the stage and quality of each tissue sample. Bjorklund et al., (2003) Lancet Neurol 2:437-45. Another contributing factor may be the host's inflammatory response to the graft. Ibid.

[0006] Another approach is to use stem cell-derived cells, such as pluripotent stem cells (PSCs), as the cell source for applications in regenerative medicine. Pluripotent stem cells can self-renew and generate all the cells needed for body tissues. PSCs comprise two main categories of cells: embryonic stem (ES) cells and induced pluripotent stem cells (iPSCs). ES cells are derived from the inner cell mass of the pre-implantation embryo and can be maintained indefinitely and expanded in vitro in their pluripotent state. (Romito and Cobellis (2016)) Stem Cells Int 2016:9451492. Recently, preliminary results from a Phase I clinical trial involving the implantation of dopaminergic neurons derived from ES cell differentiation into the brains of Parkinson's disease patients (2023 International Conference on Parkinson's Disease and Movement Disorders, Copenhagen, Denmark, August 27-31). Results showed that the strategy was well-tolerated with no serious treatment-related adverse effects. Preliminary efficacy data indicated improvements in motor function. Despite these advances, the use of embryonic stem cells remains hampered by ethical concerns and the potential for such cells to form tumors in patients. Finally, in the case of allogeneic stem cell transplantation, ES cell-derived grafts may induce immune responses in patients.

[0007] Using induced pluripotent stem cells (iPSCs) instead of ES-derived cells offers the advantage of avoiding ethical issues. Furthermore, deriving iPSCs from the patient to be treated (i.e., the patient receiving an autologous cell transplant) avoids the immune rejection risks inherent in the use of embryonic stem cells. iPSCs can be obtained by reprogramming (“dedifferentiating”) adult somatic cells into cells that are more like ES cells, including those with the ability to expand indefinitely and differentiate into all three germ layers. Ibid. Such reprogramming is typically accomplished using “Yamano factors” (Oct 3 / 4, Sox2, Klf4, and Myc family members). See, for example, U.S. Patent 8,530,238.

[0008] Various methods for differentiating pluripotent stem cells into lineage-specific cell populations and the resulting cell compositions are available for cell replacement therapy in patients with diseases leading to the loss of function of the targeted cell populations. However, in some cases, such methods are limited in their ability to produce cells with consistent physiological characteristics, and the cells produced by such methods may be limited in their ability to be transplanted in vivo and to neurally innervate other cells. For example, neural cells differentiated from pluripotent stem cells may be more readily transplanted into the brain of a treated subject when nerve cells are in an intermediate stage between an early stage (e.g., the progenitor or precursor cell stage) and a late stage (e.g., the differentiated cell stage). Moreover, there is a need to improve the manufacturability of lineage-specific cell populations derived from pluripotent stem cells, for example, for therapeutic purposes, by reducing the time and / or resources required for such manufacturing, including costs.

[0009] While many differentiation protocols can generate cells expressing dopaminergic markers, identifying suitable cells for implantation in subjects remains a challenge. For example, there is the expectation that the cells will successfully transplant and survive in the subject's brain after implantation, and that they will produce dopamine after successful implantation and survival. In some cases, regulatory agencies such as the U.S. Food and Drug Administration (FDA) and the European Medicines Agency (EMEA) require cell therapy products to undergo potency testing before approval. For instance, the FDA states that potency testing should be based on the product's mechanism of action (MOA) (Draft Guidance for Industry - Potency Assurance for Cellular and Gene Therapy Products, US Food and Drug Administration (December 2023) https: / / www.fda.gov / regulatory-information / search-fda-guidance-documents / potency-assurance-cellular-and-gene-therapy-products). For cell therapies with complex mechanistic profiles, a potency assay matrix is ​​needed that can quantitatively and accurately characterize various important cell functions after cell maturation.

[0010] Therefore, there is a need for methods to evaluate the efficacy of neuronal progenitor cells (such as those derived from stem cell differentiation) in the treatment of neurodegenerative diseases. Among these efficacy measurements, methods are needed to predict whether a population of neuronal progenitor cells is likely to successfully transplant and survive after implantation into the brain of a subject, and to predict whether neurons differentiated from the neuronal progenitor cell population are likely to produce dopamine. This invention addresses these and other needs. Summary of the Invention

[0011] In some embodiments, the present invention provides a method for predicting whether a population of neuronal progenitor cells is likely to successfully transplant and survive when implanted into a brain region. In some embodiments, these methods may include: (a) determining the gene expression levels of one or more genes (G genes) in a test sample comprising a population of neuronal progenitor cells that are associated with the predicted transplant survival potential, wherein the one or more G genes are selected from the group consisting of: AC000120.3, KRT77, TTR, PRR16, MEGF10, PDE3A, GDPD2, CMTM8, APOA1, CMTM7, CDHR3, CORIN, VTN, CPNE8, EFEMP1, CD47, SPARC, JAM2, CDO1, PLXDC2, DYNLL2, ITGA3, RPS6KL1, CHRNB2, SULT4A1, PTPN3, LZTS1, ... The G genes are: RUNX1T1, TMEM145, EPHA10, CARMIL3, MANEAL, TMEM176B, MPP3, DRAXIN, ADGRB1, KIF26A, CELF5, CNTN2, ASPHD1, SVOP, ANGPT2, SLC22A15, SRRM3, GRIN2D, DACH2, CHST1, GRIN1, LHX5, and NOS2; and (b) predicting neuronal transplant viability of neuronal progenitor cells by correlating the determined gene expression levels of the one or more G genes in the test sample with a reference map of each G gene, which correlates graft size with the gene expression levels of G genes in a training set including one or more reference samples. In some embodiments, the one or more G genes are selected from the group consisting of: TTR, PRR16, CMTM8, APOA1, CD47, CD01, KIR26A, and CNTN2. In some embodiments, the one or more G genes are selected from the group consisting of TTR, PRR16, and CD47. In some embodiments, the one or more G genes are TTR, PRR16, and CD47. In some embodiments, the expression levels of at least 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, or 20 G genes are determined.

[0012] In some embodiments, the reference plot includes one or more data points, wherein each data point on the reference plot is determined by the following steps: (a) measuring the gene expression level of the G gene in a reference sample comprising a population of neuronal progenitor cells; (b) implanting neuronal progenitor cells from the reference sample into a brain region of a test animal and measuring the size of the graft formed by the implanted neuronal progenitor cells after a culture period; and (c) plotting the graft size against the expression level of the G gene to obtain a data point for a training sample. In some embodiments, the reference plot includes multiple data points obtained for each of a plurality of reference samples. In some embodiments, the reference plot is obtained through differential expression analysis or linear regression analysis of the plurality of data points.

[0013] In some implementations, a method for predicting whether a population of neuronal progenitor cells is likely to successfully transplant and survive when implanted into a brain region involves obtaining a reference map by applying the gene expression levels of one or more G genes in a test sample as input to a machine learning model configured to predict whether neurons derived from the neuronal progenitor cells are likely to transplant and survive in the brain region after the population of neuronal progenitor cells is implanted into a subject's brain region, wherein the machine learning model is trained using the gene expression levels of G genes in multiple reference populations of neuronal progenitor cells. In some implementations, the machine learning includes principal component analysis.

[0014] In some embodiments, the predicted graft size is expressed as the number of neurons derived from neuronal progenitor cells within a cross-section of a brain region in the test animal. In some embodiments, the brain region is the substantia nigra, and in some embodiments, the cross-section of the substantia nigra includes approximately one-sixth of the substantia nigra. In some embodiments, a high graft viability is indicated if the predicted graft size is greater than 1,000 cells in one-sixth of the substantia nigra cross-section.

[0015] In some embodiments, the predicted transplant viability is determined for two or more G genes, and the overall transplant viability prediction for the test sample is based on a combined assessment of the predicted transplant viability obtained for each of the two or more G genes. In some embodiments, the combined assessment includes determining the mean or median predicted transplant viability for each of the two or more G genes.

[0016] In some embodiments, the neuronal progenitor cell population is derived from cell cultures differentiated from these cells under conditions that induce neural differentiation of pluripotent stem cells. In some embodiments, the pluripotent stem cells are induced pluripotent stem cells (iPSCs). In some embodiments, the pluripotent stem cells are autologous to the subject to be implanted with neuronal progenitor cells. In some embodiments, differentiation conditions include adherent cell culture, and neuronal progenitor cell test samples are obtained between day 18 and day 24 after the start of the differentiation process. In some embodiments, differentiation conditions include suspension cell culture, and neuronal progenitor cell test samples are obtained between day 13 and day 20 after the start of the differentiation process. In some embodiments, neuronal progenitor cell test samples are obtained at approximately day 16 after the start of the differentiation process. In some embodiments, the reference sample includes a pooled sample of neuronal progenitor cells derived from multiple donors.

[0017] In some embodiments, the gene expression level of each of the at least one G gene is determined by RNA sequencing (RNAseq). In some embodiments, the gene expression level of each of the at least one G gene is determined by polymerase chain reaction (PCR), and in some embodiments, the PCR is quantitative PCR (qPCR). In some embodiments, the gene expression level of the one or more G genes is determined by the following steps: (a) obtaining an RNA sample from a neuronal progenitor cell test sample; (b) synthesizing complementary DNA from the RNA sample using reverse transcription; (c) amplifying a specific nucleic acid fragment corresponding to a G gene using quantitative polymerase chain reaction (qPCR), wherein the qPCR includes using a pair of primers specific to the G gene and optionally a probe specific to the G gene; and (d) determining the expression level of the G gene based on a normalized quantitative amount.

[0018] In some implementations, the gene expression levels of one or more G genes are normalized to the ratio of the relative expression levels of G genes and housekeeping genes. A suitable housekeeping gene is GAPDH.

[0019] In some embodiments, the present invention provides a method for predicting the neuronal transplant viability of neurons derived from neuronal progenitor cells after implantation into the brain with a test population of neuronal progenitor cells, the method comprising: (a) determining the gene expression levels of one or more genes (G genes) associated with transplant viability in a test sample comprising the neuronal progenitor cell population, wherein the one or more G genes are selected from the group consisting of: AC000120.3, KRT77, TTR, PRR16, MEGF10, PDE3A, GDPD2, CMTM8, APOA1, CMTM7, CDHR3, CORIN, VTN, CPNE8, EFEMP1, CD47, SPARC, JAM2, CDO1, PLXDC2, DYNLL2, ITGA3, RPS6KL1, CHRNB2, SULT4A1, PTPN3, LZTS1, RUNX1T1 The expression levels of one or more G genes in the tested neuronal progenitor cell population are: (i) above the predetermined threshold for the G gene; or (ii) below the predetermined threshold for the G gene, wherein “above” or “below” is defined by the known biological relevance of the G gene to transplantation viability.

[0020] In some embodiments, a predetermined threshold for a specific G gene is based on the expression level of the G gene in a training sample, which includes neuronal progenitor cells known to exhibit high graft survival rates when implanted into the brain, and the gene expression level of the G gene in a test sample, which is similar to the expression level of the G gene in the training sample, predicts that neurons derived from the neuronal progenitor cells in the test sample have high graft survival potential. In some embodiments, a predetermined threshold for a specific G gene is based on the expression level of the G gene in a training sample, which includes neuronal progenitor cells known to exhibit low graft survival rates when implanted into the brain, and the gene expression level of the G gene in a test sample, which is similar to the expression level of the G gene in a control sample, predicts that neurons derived from the neuronal progenitor cells in the test sample have low graft survival potential.

[0021] In some implementations, high transplant viability of neuronal progenitor cells after implantation into the brain is predicted if the following conditions are met: (a) the gene expression level of at least one first G gene selected from the group consisting of: AC000120.3, KRT77, TTR, PRR16, MEGF10, PDE3A, GDPD2, CMTM8, APOA1, CMTM7, CDHR3, CORIN, VTN, CPNE8, EFEMP1, CD47, SPARC, JAM2, CDO1, and PLXDC2; and / or (b) at least one gene expression level ... CDHR8, VTN, CPNE8, EFEMP1, CD47, SPARC, JAM2, CDO1, and PLXDC2; and / or (b) at least one gene expression level selected from the group consisting of: AC000120.3, KRT77, TTR, PRR16, MEGF17, PRR18, PRR19, PRR19, PRR19, PRR19, PRR19, PRR19, PRR19, PRR19, PRR19, PRR19, PRR19, PRR19, PRR19, PRR19, PRR19, PRR19, PRR19, PRR The expression level of a second G gene is higher than a predetermined threshold for the second G gene: DYNLL2, ITGA3, RPS6KL1, CHRNB2, SULT4A1, PTPN3, LZTS1, RUNX1T1, TMEM145, EPHA10, CARMIL3, MANEAL, TMEM176B, MPP3, DRAXIN, ADGRB1, KIF26A, CELF5, CNTN2, ASPHD1, SVOP, ANGPT2, SLC22A15, SRRM3, GRIN2D, DACH2, CHST1, GRIN1, LHX5, and NOS2.

[0022] In some implementations, a predetermined threshold for a specific G gene is based on the ratio of the relative expression levels of a) the G gene and b) the control gene in the test sample. In some implementations, the control gene is GAPDH, and the pre-determined thresholds are selected from a group consisting of: (a) a ratio of AC000120.3 to GAPDH expression less than about 0.14; (b) a ratio of KRT77 to GAPDH expression less than about 0.68; (c) a ratio of TTR to GAPDH expression less than about 1.11; (d) a ratio of PRR16 to GAPDH expression less than about 0.43; (e) a ratio of MEGF10 to GAPDH expression less than about 0.79; (f) a ratio of PDE3A to GAPDH expression less than about 1.00; (g) a ratio of GDPD2 to GAPDH expression less than about 0.78; (h) a ratio of CMTM8 to GAPDH expression less than about 1.02; (i) a ratio of APOA1 to GAPDH expression less than about 0.68; (j) a ratio of CMTM7 to GAPDH expression less than about 0.88; (k) ... The following ratios were observed: (l) CDHR3 to GAPDH expression ratio less than approximately 1.09; (m) CORIN to GAPDH expression ratio less than approximately 1.24; (n) VTN to GAPDH expression ratio less than approximately 0.98; (o) CPNE8 to GAPDH expression ratio less than approximately 0.79; (p) EFEMP1 to GAPDH expression ratio less than approximately 0.83; (q) CD47 to GAPDH expression ratio less than approximately 1.16; (r) SPARC to GAPDH expression ratio less than approximately 1.29; (s) JAM2 to GAPDH expression ratio less than approximately 0.82; (t) CDO1 to GAPDH expression ratio less than approximately 1.00; (v) PLXDC2 to GAPDH expression ratio less than approximately 1.00; (u) DYNLL2 to GAPDH expression ratio greater than approximately 0.56; (v) The following are examples of protein expression ratios: (w) ITGA3 to GAPDH expression ratio greater than approximately 0.26; (x) RPS6KL1 to GAPDH expression ratio greater than approximately 0.21; (y) CHRNB2 to GAPDH expression ratio greater than approximately 0.23; (y) SULT4A1 to GAPDH expression ratio greater than approximately 0.22; (z) PTPN3 to GAPDH expression ratio greater than approximately 0.03; (aa) LZTS1 to GAPDH expression ratio greater than approximately 0.19; (ab) RUNX1T1 to GAPDH expression ratio greater than approximately 0.24; (ac) TMEM145 to GAPDH expression ratio greater than approximately 0.05; (ad) EPHA10 to GAPDH expression ratio greater than approximately 0.16; (ae) CARMIL3 to GAPDH expression ratio greater than approximately 0.The ratios of 16, (af) MANEAL to GAPDH expression greater than approximately 0.24, (ag) TMEM176B to GAPDH expression greater than approximately 0.11, (ah) MPP3 to GAPDH expression greater than approximately 0.12, (ai) DRAXIN to GAPDH expression greater than approximately 0.27, (aj) ADGRB1 to GAPDH expression greater than approximately 0.07, (ak) KIF26A to GAPDH expression greater than approximately 0.23, (al) CELF5 to GAPDH expression greater than approximately 0.25, (am) CNTN2 to GAPDH expression greater than approximately 0.23, (an) ASPHD1 to GAPDH expression greater than approximately 0.08, (ao) SVOP to GAPDH expression greater than approximately 0.16, (ap) The ratios of ANGPT2 to GAPDH expression were greater than approximately 0.06; (aq) the ratio of SLC22A15 to GAPDH expression was greater than approximately 0.04; (ar) the ratio of SRRM3 to GAPDH expression was greater than approximately 0.17; (as) the ratio of GRIN2D to GAPDH expression was greater than approximately 0.02; (at) the ratio of DACH2 to GAPDH expression was greater than approximately 0.06; (au) the ratio of CHST1 to GAPDH expression was greater than approximately 0.04; (av) the ratio of GRIN1 to GAPDH expression was greater than approximately 0.26; (aw) the ratio of LHX5 to GAPDH expression was greater than approximately 0.06; and (ax) the ratio of NOS2 to GAPDH expression was greater than approximately 0.08.

[0023] In some embodiments, the present invention also provides methods for training a machine learning model to predict whether a population of neuronal progenitor cells is likely to successfully transplant and survive when implanted into a brain region. These methods may include: (a) obtaining gene expression levels of one or more genes in each of a plurality of reference populations of neuronal progenitor cells; (b) receiving transplantation survival adaptation information for each of the plurality of reference populations, wherein the transplantation survival adaptation information of the reference populations indicates whether or to what extent the neuronal progenitor cells have transplanted and survived in the brain region after implantation of the reference populations of neuronal progenitor cells into a brain region of a subject; and (c) applying the gene expression levels of (a) and the transplantation survival adaptation information of (b) as inputs to train a machine learning model, wherein the machine learning model is trained to predict, based on the gene expression levels of multiple genes, whether the neuronal progenitor cell population will transplant and survive in the brain region after implantation into a brain region of a subject.

[0024] In some embodiments, the present invention provides a computing device configured to predict the transplant viability potential of neuronal progenitor cells when a population of neuronal progenitor cells is implanted into a brain region, the computing device comprising: (a) a processor; and (b) a memory including instructions executable by the processor, the instructions being configured to perform the following steps: (i) receiving a test sample comprising gene expression data of one or more genes (G genes) in the neuronal progenitor cell population associated with the predicted transplant viability potential, wherein the one or more G genes are selected The following groups are free: AC000120.3, KRT77, TTR, PRR16, MEGF10, PDE3A, GDPD2, CMTM8, APOA1, CMTM7, CDHR3, CORIN, VTN, CPNE8, EFEMP1, CD47, SPARC, JAM2, CDO1, PLXDC2, DYNLL2, ITGA3, RPS6KL1, CHRNB2, SULT4A1, PTPN3, LZTS1, RUNX1T1, TMEM 145, EPHA10, CARMIL3, MANEAL, TMEM176B, MPP3, DRAXIN, ADDRB1, KIF26A, CELF5, CNTN2, ASPHD1, SVOP, ANGPT2, SLC22A15, SRRM3, GRIN2D, DACH2, CHST1, GRIN1, LHX5, and NOS2; (ii) determining the gene expression level of each of the one or more G genes based on the test sample; (iii) comparing the determined gene expression level of each of the one or more G genes in the test sample with a reference plot for each corresponding G gene, wherein each reference plot correlates the gene expression level of the G gene with graft size data obtained from a training set including one or more reference samples; and (iv) generating a predictive assessment of the transplant survival potential of a neuronal progenitor cell population by predicting the neuronal transplant survival ability of neuronal progenitor cells in the test sample by correlating the determined gene expression level of the one or more G genes in the test sample with reference plot data.

[0025] In some embodiments, the present invention also provides a kit for predicting the transplant viability potential of neurons derived from neuronal progenitor cells. Such a kit may include one or more of the following: (a) a first pair of oligonucleotide primers suitable for amplifying a first gene; (b) a second pair of oligonucleotide primers suitable for amplifying a second gene; and (c) a third pair of oligonucleotide primers suitable for amplifying a third gene; wherein each of the first, second, and third genes is selected from the group consisting of: AC000120.3, KRT77, TTR, PRR16, MEGF10, PDE3A, GDPD2, CMTM8, APOA1, CMTM7, CDHR3, CORIN, VTN, CPNE8, EFEMP1, CD47, SPARC. JAM2, CDO1, PLXDC2, DYNLL2, ITGA3, RPS6KL1, CHRNB2, SULT4A1, PTPN3, LZTS1, RUNX1T1, TMEM145, EPHA10, CARMIL3, MANEAL, TMEM176B, MPP3, DRAXIN, ADGRB1, KIF26A, CELF5, CNTN2, ASPHD1, SVOP, ANGPT2, SLC22A15, SRRM3, GRIN2D, DACH2, CHST1, GRIN1, LHX5, and NOS2. In some implementations, the first, second, and third genes are each selected from the group consisting of: TTR, PRR16, CMTM8, APOA1, CD47, CD01, KIR26A, and CNTN2. In some embodiments, the kit includes at least three pairs of oligonucleotide primers, with the first gene being TTR, the second gene being PRR16, and the third gene being CD47. In some embodiments, the expression levels of at least 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, or 20 D genes are determined. In some embodiments, the reference sample includes a pooled sample of neuronal progenitor cells derived from multiple donors.

[0026] In some embodiments, the present invention provides a method for predicting whether neurons derived from a population of neuronal progenitor cells will produce dopamine. In some embodiments, these methods involve: (a) determining the gene expression levels of one or more genes (D genes) associated with predicted dopamine production in a test sample comprising a population of neuronal progenitor cells, wherein the D genes are selected from the group consisting of: CNTNAP5, KLHL1, NHLH2, GREM2, BRINP2, GRIN3A, LRRC4C, IRX3, CPNE4, PTPN3, PMEL, PCDH20, LRRC37A2, TMEM246, B3GALNT1, ZHX1, BCAS4, SLC25A37, GRINA, MID1, FRMD4A, PARP10, WHAMMP2, EYA1, CORO2B, WHAMMP3, B3GALT5, GPR35, ABCD2, ITI H3, AC107464.1, CAMK2N1, CAMK2A, PRPS1, GOLGA6L10, AMOT, SULT1A1, CD83, SPON1, FRMPD3, AC096570.1, TCAF2, GOLGA8M, VWA5B2, CA8, AC017050.1, KRT77, AP000350.6, LINC02751, and ARHGAP5-AS1; and (b) predicting the dopamine production capacity of neurons derived from neuronal progenitor cells by associating the determined gene expression levels of the one or more D genes in the test samples with a reference map of each D gene, the reference map associating neuronal dopamine production with the gene expression levels of the D genes in a training set including one or more reference samples. In some embodiments, the one or more D genes are selected from the group consisting of: CNTNAP5, NHLH2, GREM2, PMEL, PCDH20, LRRC37A2, SLC25A37, MID1, EYA1, B3GALT5, GPR35, AC107464.1, CAMK2N1, CAMK2A, GOLGA6L10, FRMPD3, VWA5B2, AC017050.1, and LINC02751. In some embodiments, the one or more D genes are selected from the group consisting of: B3GALT5, FRMPD3, and GREM2. In some embodiments, the one or more D genes are B3GALT5, FRMPD3, and GREM2.

[0027] In some embodiments, each data point on the reference plot is determined by the following steps: (a) measuring the gene expression level of the D gene in a reference sample comprising a population of neuronal progenitor cells; (b) differentiating the neuronal progenitor cells to generate neurons and measuring the amount of dopamine produced by the neurons derived from the neuronal progenitor cells; and (c) plotting dopamine production against the expression level of the D gene to obtain data points for training samples. In some embodiments, the reference plot includes multiple data points obtained for each of a plurality of reference samples. In some embodiments, the reference plot is obtained through differential expression analysis or linear regression analysis of the plurality of data points.

[0028] In some implementations, the reference plot is obtained by applying the gene expression levels of one or more D genes in the test sample as input to a machine learning model configured to predict whether neurons derived from a population of neuronal progenitor cells will produce dopamine, wherein the machine learning model is trained using the gene expression levels of the D gene in multiple reference populations of neuronal progenitor cells. In some implementations, the machine learning model includes principal component analysis.

[0029] In some implementations, if the predicted amount of dopamine produced by neurons derived from neuronal progenitor cells is at least 15 nM dopamine / 10 5 If a population of neurons is identified, it is predicted that the neuronal progenitor cell population will have a high dopamine production capacity.

[0030] In some embodiments, the predicted dopamine production capacity is determined for two or more D genes, and the overall dopamine production capacity prediction for the test sample is based on a combined assessment of the predicted dopamine production capacity for each of the two or more D genes. In some embodiments, the combined assessment includes determining mean or median predicted transplant viability.

[0031] In some embodiments, the present invention provides a method for predicting whether neurons derived from a population of neuronal progenitor cells will produce dopamine, the method comprising: (a) determining the gene expression levels of one or more genes (D genes) in a test sample comprising a population of neuronal progenitor cells that are associated with the predicted dopamine-producing capacity, wherein the one or more D genes are selected from the group consisting of: CNTNAP5, KLHL1, NHLH2, GREM2, BRINP2, GRIN3A, LRRC4C, IRX3, CPNE4, PTPN3, PMEL, PCDH20, LRRC37A2, TMEM246, B3GALNT1, ZHX1, BCAS4, SLC25A37, GRINA, MID1, FRMD4A, PA RP10, WHAMMP2, EYA1, CORO2B, WHAMMP3, B3GALT5, GPR35, ABCD2, ITIH3, AC107464.1, CAMK2N1, CAMK2A, PRPS1, GOLGA6L10, AMOT, SULT1A1, CD83, SPON1, FRMPD3, AC096570.1, TCAF2, GOLGA8M, VWA5B2, CA8, AC017050.1, KRT77, AP000350.6, LINC02751, and ARHGAP5-AS1; and (b) comparing the expression level of each of the one or more D genes in the tested neuronal progenitor cell population with a predetermined threshold for the specific D gene. The expression level indicates that the neuronal progenitor has a high predicted dopamine production capacity if it is either (i) above a predetermined threshold for the D gene or (ii) below a predetermined threshold for the D gene, wherein “above” or “below” is defined by the known biological relevance of the G gene to the predicted dopamine production capacity.

[0032] In some embodiments, a predetermined threshold for a specific D gene is based on the expression level of the D gene in training samples, which include neuronal progenitor cells known to generate neurons producing high levels of dopamine, and the gene expression level of the D gene in test samples with similar expression levels to those in the training samples predicts that neurons derived from the neuronal progenitor cells in the test samples have a high dopamine-producing potential. In some embodiments, the predetermined threshold for a specific D gene is based on the expression level of the D gene in training samples, which include neuronal progenitor cells known to generate neurons producing low levels of dopamine, and the gene expression level of the D gene in test samples with similar expression levels to those in control samples predicts that neurons derived from the neuronal progenitor cells in the test samples have a low dopamine-producing potential.

[0033] In some implementations, neuronal progenitors are predicted to produce neurons with high dopamine production capacity if the following conditions are met: (a) the gene expression level of at least one first D gene selected from the group consisting of: CNTNAP5, KLHL1, NHLH2, GREM2, BRINP2, GRIN3A, LRRC4C, IRX3, CPNE4, PTPN3, PMEL, PCDH20, LRRC37A2, TMEM246, B3GALNT1, and ZHX1; and / or (b) the gene expression level of at least one second D gene selected from the group consisting of: B CAS4, SLC25A37, GRINA, MID1, FRMD4A, PARP10, WHAMMP2, EYA1, CORO2B, WHAMMP3, B3GALT5, GPR35, ABCD2, ITIH3, AC107464.1, CAMK2N1, CAMK2A, PRPS1, G OLGA6L10, AMOT, SULT1A1, CD83, SPON1, FRMPD3, AC096570.1, TCAF2, GOLGA8M, VWA5B2, CA8, AC017050.1, KRT77, AP000350.6, LINC02751 and ARHGAP5-AS1.

[0034] In some implementations, a predetermined threshold for a specific D gene is based on the ratio of the relative expression levels of a) the D gene and b) the control gene in the test sample. In some implementations, the control gene is GAPDH, and the pre-determined thresholds are selected from a group consisting of: (a) a ratio of CNTNAP5 to GAPDH expression less than about 0.12; (b) a ratio of KLHL1 to GAPDH expression less than about 0.10; (c) a ratio of NHLH2 to GAPDH expression less than about 0.56; (d) a ratio of GREM2 to GAPDH expression less than about 0.35; (e) a ratio of BRINP2 to GAPDH expression less than about 0.97; (f) a ratio of GRIN3A to GAPDH expression less than about 0.48; (g) a ratio of LRRC4C to GAPDH expression less than about 0.39; (h) a ratio of IRX3 to GAPDH expression less than about 0.55; (i) a ratio of CPNE4 to GAPDH expression less than about 0.28; (j) a ratio of PTPN3 to GAPDH expression less than about 0.25; (k) a ratio of PTPN3 to GAPDH expression less than about 0.25; The following ratios were observed: (l) PMEL to GAPDH expression ratio less than approximately 0.29; (m) PCDH20 to GAPDH expression ratio less than approximately 0.20; (n) LRRC37A2 to GAPDH expression ratio less than approximately 0.68; (n) TMEM246 to GAPDH expression ratio less than approximately 0.53; (o) B3GALNT1 to GAPDH expression ratio less than approximately 0.67; (p) ZHX1 to GAPDH expression ratio less than approximately 0.55; (q) BCAS4 to GAPDH expression ratio greater than approximately 0.42; (r) SLC25A37 to GAPDH expression ratio greater than approximately 0.38; (s) GRINA to GAPDH expression ratio greater than approximately 0.60; (t) MID1 to GAPDH expression ratio greater than approximately 0.62; (u) The ratio of FRMD4A to GAPDH expression greater than approximately 0.57; (v) the ratio of PARP10 to GAPDH expression greater than approximately 0.25; (w) the ratio of WHAMMP2 to GAPDH expression greater than approximately 0.37; (x) the ratio of EYA1 to GAPDH expression greater than approximately 0.32; (y) the ratio of CORO2B to GAPDH expression greater than approximately 0.40; (z) the ratio of WHAMMP3 to GAPDH expression greater than approximately 0.34; (aa) the ratio of B3GALT5 to GAPDH expression greater than approximately 0.40; (ab) the ratio of GPR35 to GAPDH expression greater than approximately 0.19; (ac) the ratio of ABCD2 to GAPDH expression greater than approximately 0.35; (ad) the ratio of ITIH3 to GAPDH expression greater than approximately 0.17; (ae) AC107464.The ratios of 1 to GAPDH expression greater than approximately 0.20; (af) CAMK2N1 to GAPDH expression greater than approximately 0.52; (ag) CAMK2A to GAPDH expression greater than approximately 0.37; (ah) PRPS1 to GAPDH expression greater than approximately 0.52; (ai) GOLGA6L10 to GAPDH expression greater than approximately 0.21; (aj) AMOT to GAPDH expression greater than approximately 0.50; (ak) SULT1A1 to GAPDH expression greater than approximately 0.18; (al) CD83 to GAPDH expression greater than approximately 0.29; (am) SPON1 to GAPDH expression greater than approximately 0.76; (an) FRMPD3 to GAPDH expression greater than approximately 0.31; (ao) The ratio of AC096570.1 to GAPDH expression greater than approximately 0.14; (ap) TCAF2 to GAPDH expression greater than approximately 0.30; (aq) GOLGA8M to GAPDH expression greater than approximately 0.003; (ar) VWA5B2 to GAPDH expression greater than approximately 0.22; (as) CA8 to GAPDH expression greater than approximately 0.19; (at) AC017050.1 to GAPDH expression greater than approximately 0.08; (au) KRT77 to GAPDH expression greater than approximately 0.14; (av) AP000350.6 to GAPDH expression greater than approximately 0.31; (aw) LINC02751 to GAPDH expression greater than approximately 0.19; and (ax) ARHGAP5-AS1 to GAPDH expression greater than approximately 0.26.

[0035] In some embodiments, the present invention provides methods for training machine learning models to predict whether neurons derived from a population of neuronal progenitor cells will produce dopamine. These methods may include: (a) obtaining gene expression levels of one or more genes in each of a plurality of reference populations of neuronal progenitor cells; (b) receiving dopamine production information of neurons derived from each of the plurality of reference populations, wherein the dopamine production information of the reference populations indicates whether or to what extent cells derived from neuronal progenitor cells produce dopamine; and (c) applying the gene expression levels of (a) and the dopamine production information of (b) as input to train a machine learning model, wherein the machine learning model is trained to predict whether neurons derived from a population of neuronal progenitor cells will produce dopamine based on the gene expression levels of a plurality of genes.

[0036] In some embodiments, the present invention provides a kit for predicting dopamine production in neurons derived from a population of neuronal progenitor cells. The kit of the present invention comprises one or more of the following: (a) a first pair of oligonucleotide primers suitable for amplifying a first gene; (b) a second pair of oligonucleotide primers suitable for amplifying a second gene; and (c) a third pair of oligonucleotide primers suitable for amplifying a third gene; wherein each of the first, second, and third genes is selected from the group consisting of: CNTNAP5, KLHL1, NHLH2, GREM2, BRINP2, GRIN3A, LRRC4C, IRX3, CPNE4, PTPN3, PMEL, PCDH20, LRRC37A2, TMEM246, B3GALNT1, ZHX1, BCAS4, SLC25A37 , GRINA, MID1, FRMD4A, PARP10, WHAMMP2, EYA1, CORO2B, WHAMMP3, B3GALT5, GPR35, ABCD2, ITIH3, AC107464.1, CAMK2N1, CAMK2A, PRPS1, GOLGA6L 10. AMOT, SULT1A1, CD83, SPON1, FRMPD3, AC096570.1, TCAF2, GOLGA8M, VWA5B2, CA8, AC017050.1, KRT77, AP000350.6, LINC02751 and ARHGAP5-AS1. In some embodiments, the first, second, and third genes are each selected from the group consisting of: CNTNAP5, NHLH2, GREM2, PMEL, PCDH20, LRRC37A2, SLC25A37, MID1, EYA1, B3GALT5, GPR35, AC107464.1, CAMK2N1, CAMK2A, GOLGA6L10, FRMPD3, VWA5B2, AC017050.1, and LINC02751. In some embodiments, the first gene is B3GALT5, the second gene is GREM2, and the third gene is FRMPD3.

[0037] In some embodiments, the present invention provides a computing device configured to predict whether neurons differentiating from a population of neuronal progenitor cells will produce dopamine. The computing device includes: (a) a processor; and (b) a memory including instructions executable by the processor, the instructions being configured to perform the following steps: (i) receiving a test sample comprising gene expression data of one or more genes (D genes) in the neuronal progenitor cell population associated with predicted dopamine production potential, wherein the D genes are selected from the group consisting of: CNTNAP5, KLHL1, NHLH2, GREM2, BRINP2, GRIN3A, LRRC4C, IRX3. , CPNE4, PTPN3, PMEL, PCDH20, LRRC37A2, TMEM246, B3GALNT1, ZHX1, BCAS4, SLC25A37, GRINA, MID1, FRMD4A, PAR P10, WHAMMP2, EYA1, CORO2B, WHAMMP3, B3GALT5, GPR35, ABCD2, ITIH3, AC107464.1, CAMK2N1, CAMK2A, PRPS1, GO LGA6L10, AMOT, SULT1A1, CD83, SPON1, FRMPD3, AC096570.1, TCAF2, GOLGA8M, VWA5B2, CA8, AC017050.1, KRT77, AP000350.6, LINC02751, and ARHGAP5-AS1; (ii) determining the gene expression level of each of the one or more D genes based on the test sample; (iii) comparing the determined gene expression level of each of the one or more D genes in the test sample with a reference plot for each corresponding D gene, wherein each reference plot correlates the gene expression level of the D gene with the dopamine production level obtained from a training set including one or more reference samples; and (iv) generating a predictive assessment of the dopamine production potential of derived neurons by predicting the dopamine production capacity of neurons derived from neuronal progenitor cells in the test sample by correlating the determined gene expression level of the one or more D genes with reference plot data.

[0038] In some embodiments, the present invention provides a potency assay matrix for determining the efficacy of a neuronal progenitor cell population in treating neurodegenerative diseases. The potency assay matrix includes a method of subjecting the neuronal progenitor cell population to at least two of the steps (a), (b), and (c): (a) Classifying an in vitro population of neuronal progenitor cells to determine whether neuronal progenitor cells include identifiable dopaminergic precursor cells by the following steps: (i) receiving a test dataset as input, the test dataset including the expression levels of one or more genes expressed in a first test sample including neuronal progenitor cells; (ii) using the test dataset and a first reference dataset to compute a first similarity score for the first test sample, wherein: (1) the first reference dataset includes a representation of gene expression levels of one or more genes differentially expressed between cells in a first differentiation state and cells in a second differentiation state, wherein the second differentiation state is the differentiation state of identifiable dopaminergic neurons, and wherein (i) The first differentiation state is earlier or later than the second differentiation state in the stem cell differentiation pathway; (ii) The expression levels in the test dataset include the expression levels of one or more genes included in the first reference dataset; and (iii) The first similarity score indicates whether the differentiation state of the test cells is more similar to the first or second differentiation state; (iv) The novelty score of the neuronal progenitor cells in the first test sample is determined, wherein the novelty score indicates the degree of deviation of the gene expression levels in the test dataset from the gene expression levels in the reference database; and (v) Based on the similarity score and the novelty score, it is determined whether the first test sample includes the identified dopaminergic neuronal cells. (b) To predict the likelihood of successful transplantation survival of neuronal progenitor cells when implanted into a brain region, the following steps are performed: (i) Determine the gene expression levels of one or more genes (G genes) in a second test sample containing neuronal progenitor cells that are associated with the predicted transplantation survival potential, wherein the one or more G genes are selected from the group consisting of: AC000120.3, KRT77, TTR, PRR16, MEGF10, PDE3A, GDPD2, CMTM8, APOA1, CMTM7, CDHR3, CORIN, VTN, CPNE8, EFEMP1, CD47, SPARC, JAM2, CDO1, PLXDC2, DYNLL2, ITGA3, RPS6KL1, CHRNB2, SULT4A1, PTPN3. LZTS1, RUNX1T1, TMEM145, EPHA10, CARMIL3, MANEAL, TMEM176B, MPP3, DRAXIN, ADDRB1, KIF26A, CELF5, CNTN2, ASPHD1, SVOP, ANGPT2, SLC22A15, SRRM3, GRIN2D, DACH2, CHST1, GRIN1, LHX5, and NOS2; and (ii) predicting neuronal transplant viability of neuronal progenitor cells by correlating the determined gene expression levels of one or more G genes in a second test sample with a reference map of each G gene, the reference map correlating graft size with gene expression levels of G genes in a training set including one or more reference samples; and (c) To predict whether neurons derived from a population of neuronal progenitor cells will produce dopamine, the following steps are performed: (i) Determine the gene expression levels of one or more genes (D genes) associated with predicted dopamine production in a third test sample comprising the population of neuronal progenitor cells, wherein the D genes are selected from the group consisting of: CNTNAP5, KLHL1, NHLH2, GREM2, BRINP2, GRIN3A, LRRC4C, IRX3, CPNE4, PTPN3, PMEL, PCDH20, LRRC37A2, TMEM246, B3GALNT1, ZHX1, BCAS4, SLC25A37, GRINA, MID1, FRMD4A, PARP10, WHAMMP2, EYA1, CORO2B, WHAMMP3, B3GALT5, GPR35, ABCD. 2. ITIH3, AC107464.1, CAMK2N1, CAMK2A, PRPS1, GOLGA6L10, AMOT, SULT1A1, CD83, SPON1, FRMPD3, AC096570.1, TCAF2, GOLGA8M, VWA5B2, CA8, AC017050.1, KRT77, AP000350.6, LINC02751, and ARHGAP5-AS1; and (ii) predicting the dopamine production capacity of neurons derived from neuronal progenitors by associating the determined gene expression levels of the one or more D genes in a third test sample with a reference map of each D gene, the reference map associating dopamine production of neurons with the gene expression levels of the D genes in a training set including one or more reference samples of neuronal progenitors.

[0039] In some embodiments, the power measurement matrix includes steps (a) and (b). In some embodiments, the power measurement matrix includes steps (b) and (c). In some embodiments, the power measurement matrix includes steps (a) and (c). In some embodiments, the power measurement matrix includes all three steps of steps (a), (b), and (c).

[0040] In some embodiments, the power assay matrix includes step (b), and the G gene is selected from the group consisting of: TTR, PRR16, CMTM8, APOA1, CD47, CD01, KIR26A, and CNTN2. In some embodiments, the one or more G genes are TTR, PRR16, and CD47.

[0041] In some embodiments, the power assay matrix includes step (c), and the one or more D genes are selected from the group consisting of: CNTNAP5, NHLH2, GREM2, PMEL, PCDH20, LRRC37A2, SLC25A37, MID1, EYA1, B3GALT5, GPR35, AC107464.1, CAMK2N1, CAMK2A, GOLGA6L10, FRMPD3, VWA5B2, AC017050.1, and LINC02751. In some embodiments, the one or more D genes are B3GALT5, FRMPD3, and GREM2.

[0042] This article also provides a therapeutic composition comprising a population of neuronal progenitor cells selected by any of the methods disclosed herein.

[0043] This article also provides a therapeutic composition comprising neuronal progenitor cells derived from pluripotent stem cells, wherein the therapeutic composition comprises neuronal progenitor cells from at least two populations selected from the group consisting of: (a) a first neuronal progenitor cell population, which is classified as a defined dopaminergic precursor cell using a method including classifying neuronal progenitor cells based on probability scores and deviation scores; (b) a second neuronal progenitor cell population, which is predicted to produce neurons with high transplant viability; and (c) a third neuronal progenitor cell population, which is predicted to produce neurons with high dopamine production.

[0044] In some embodiments, the therapeutic composition comprises a pharmaceutically acceptable carrier.

[0045] In some embodiments, the therapeutic composition contains a cryoprotectant.

[0046] In some implementations, the cryoprotectant is selected from the group consisting of glycerol, propylene glycol, and dimethyl sulfoxide (DMSO).

[0047] In some embodiments, the composition is used to treat a subject with a neurodegenerative disease or condition, optionally wherein the neurodegenerative disease or condition includes the loss of dopaminergic neurons.

[0048] In some embodiments, the composition is used to manufacture a medicament for treating a subject with a neurodegenerative disease or condition, optionally wherein the neurodegenerative disease or condition includes the loss of dopaminergic neurons.

[0049] This article also provides a treatment method comprising implanting a therapeutically effective amount of any of the therapeutic compositions disclosed herein into a brain region of a subject suffering from a neurodegenerative disease or condition, optionally wherein the neurodegenerative disease or condition includes loss of dopaminergic neurons.

[0050] This article also provides a method for transplanting and surviving neuronal cells in a brain region of a subject, the method comprising implanting a therapeutically effective amount of any of the therapeutic compositions disclosed herein into a brain region of a subject suffering from a neurodegenerative disease or symptom.

[0051] This article also provides a method for increasing dopamine production in brain regions of a subject, the method comprising implanting a therapeutically effective amount of any of the therapeutic compositions disclosed herein into brain regions of a subject suffering from a neurodegenerative disease or condition.

[0052] In some embodiments, the neurodegenerative disease or condition includes the loss of dopaminergic neurons in the substantia nigra, optionally in the substantia nigra sclerosis (SNc). In some embodiments, the neurodegenerative disease or condition is Parkinson's disease. In some embodiments, the neurodegenerative disease or condition is Parkinson's syndrome. In some embodiments, the brain region is the substantia nigra.

[0053] In some embodiments, the implantation is performed via stereotactic injection. In some embodiments, the neuronal progenitor cells of the therapeutic composition are autologous to the subject. Attached Figure Description

[0054] Various aspects of the disclosed methods, apparatus, and systems are specifically set forth in the appended claims. A better understanding of the features and advantages of the disclosed methods, apparatus, and systems will be obtained by referring to the following detailed description and accompanying drawings of exemplary embodiments, wherein: Figure 1 Non-limiting exemplary methods for predicting whether a population of neuronal progenitor cells is likely to successfully transplant and survive when implanted into a brain region are described according to some embodiments of the present disclosure.

[0055] Figure 2 Non-limiting exemplary methods for predicting whether neurons derived from neuronal progenitor cells will produce dopamine after implantation of a population of neuronal progenitor cells into a brain region are described according to some embodiments of the present disclosure.

[0056] Figure 3 A non-limiting exemplary schematic diagram is depicted for generating neuronal progenitor cells from a subject for transplantation and survival in a rodent brain and for quantifying gene expression levels.

[0057] Figure 4Exemplary computing devices or systems according to some embodiments of this disclosure are depicted.

[0058] Figure 5 Exemplary computer systems or computer networks are depicted, illustrating some instances of the systems described herein.

[0059] Figure 6 A non-limiting exemplary schematic diagram is depicted for deriving a linear regression model from principal component analysis (PCA) to predict the size of a neuronal cell graft derived from a neuronal progenitor cell after implantation into a brain region of a subject, based on the gene expression level of the neuronal progenitor cell.

[0060] Figure 7 Non-limiting examples of the data are depicted, showing that upregulation of expression of five genes (genes 1 to 5) corresponds to larger predicted graft size, and upregulation of expression of five genes (genes 6 to 10) corresponds to smaller predicted graft size.

[0061] Figure 8 A to Figure 8 H depicts an unrestricted example of data that illustrates linear regression models derived from PCA. These models predict the size of grafts obtained from neuronal progenitor cells when they are implanted into a subject's brain, based on gene expression levels in a population of neuronal progenitor cells. These models are based on their R... 2 The values ​​are organized by the percentage of variance explained by PC1 (PC1%).

[0062] Figure 9 Non-limiting examples of data are depicted, illustrating the predictive performance of linear regression models derived from PCA, which predict the size of grafts obtained from neuronal progenitor cells when they are implanted into the brains of subjects, based on the gene expression levels of neuronal progenitor cell populations.

[0063] Figure 10 Unrestricted examples of data are depicted, illustrating the summary statistics of linear regression models derived from PCA that predict the size of grafts obtained from neuronal progenitor cells when they are implanted into the brains of subjects, based on gene expression levels of neuronal progenitor cell populations.

[0064] Figure 11 Non-limiting exemplary schematic diagrams are depicted for deriving linear regression models without using PCA, which are used to predict the size of the graft obtained from neuronal progenitor cells when they are implanted into the brain of a subject, based on the gene expression levels of a neuronal progenitor cell population.

[0065] Figure 12A to Figure 12 C depicts a non-restrictive example of a linear regression dataset derived without using PCA, which is used to predict whether a population of neuronal progenitor cells is likely to successfully transplant and survive when implanted into a brain region based on the gene expression levels of three genes.

[0066] Figure 13 A non-limiting example of data is depicted, illustrating the predictive performance of a linear regression model without the use of PCA, which predicts the size of a graft obtained by implanting neuronal progenitor cells into the brain of a subject based on gene expression levels in neuronal progenitor cells.

[0067] Figure 14 A non-limiting example of the data is depicted, illustrating the summary statistics of a linear regression model without the use of PCA, which predicts the size of a graft obtained by implanting neuronal progenitor cells into the brain of a subject based on gene expression levels in neuronal progenitor cells.

[0068] Figure 15 A to Figure 15 B depicts a non-restrictive example of data that validates the performance of a linear regression model derived without PCA, used to predict graft size obtained by implanting neuronal progenitor cells into the brain of a subject based on gene expression levels in neuronal progenitor cells. Out-of-sample data and labels are from Kirkeby et al., (2017). Cell Stem Cell 20(1): 135-148. The number of Th+ cells in the graft / 100,000 injected cells provided a basis for the “DA yield” label (high DA / low DA); n=15 predictions. All graft measurements were performed on a single cell line (H9 ESC).

[0069] Figure 16 A to Figure 16 F depicts a non-restrictive example of the data, illustrating the good agreement between measured and predicted graft sizes for six donor cell batches when a linear regression model derived without PCA is used for prediction.

[0070] Figure 17 A to Figure 17 C depicts a non-limiting example of data, illustrating an image of a graft obtained by implanting neuronal progenitor cells into the brain of a rodent host. The cell nucleus is shown.

[0071] Figure 18 A to Figure 18H depicts an unrestricted example of data that illustrates linear regression models derived from PCA. These models predict the amount of dopamine released from neurons derived from neuronal progenitor cells after implantation into the subject's brain, based on gene expression levels in a population of neuronal progenitor cells. These models are based on their R... 2 The values ​​are organized by the percentage of variance explained by PC1 (PC1%).

[0072] Figure 19 A non-limiting example of data is depicted, illustrating the predictive performance of a linear regression model derived from PCA, which predicts the amount of dopamine released by neurons derived from neuronal progenitor cells after implantation of neuronal progenitor cells into the brain of a subject, based on gene expression levels in a population of neuronal progenitor cells.

[0073] Figure 20 A to Figure 20 C depicts a non-restrictive example of a linear regression set of data derived without using PCA, which is used to predict the amount of dopamine released by neurons derived from neuronal progenitors after implantation into a subject's brain based on gene expression levels in a population of neuronal progenitors.

[0074] Figure 21 A non-limiting example of data is depicted, illustrating the predictive performance of a linear regression model derived without PCA, which predicts the amount of dopamine released by neurons derived from neuronal progenitor cells after implantation of neuronal progenitor cells into the subject's brain based on gene expression levels in a population of neuronal progenitor cells. Detailed Implementation

[0075] In some embodiments, this document provides methods for characterizing populations of neuronal progenitor cells (hereinafter also referred to as dopaminergic neuronal progenitor cells (DANPCs)), such as by predicting whether the neuronal progenitor cell population or neurons derived from such population are likely to exhibit one or more functions, activities, or differentiation states. In some embodiments, this document also provides methods for selecting populations of neuronal progenitor cells, wherein such cells have one or more desired characteristics used to predict whether such neuronal progenitor cells or neurons derived from such populations are likely to exhibit one or more functions, activities, or differentiation states. In some embodiments, this document also provides methods for implanting any such selected population of neuronal progenitor cells into a subject. In some embodiments, the one or more characteristics reflect the mechanism of action of the neurons after implantation of the neuronal progenitor cells into the subject. In some embodiments, the desired characteristic is the ability of neurons derived from the neuronal progenitor cells to survive transplantation in a brain region of the subject after implantation of the neuronal progenitor cells. In some embodiments, the desired characteristic is the ability of neurons derived from the neuronal progenitor cells to produce dopamine after implantation of the neuronal progenitor cells into the brain. In some implementations, these methods can also be used to identify or select populations of neuronal progenitor cells that have the following characteristics: differentiation state that is a defined dopaminergic neuron differentiation state.

[0076] In some embodiments, this document also provides a computing device, including a computing device for performing any of the provided methods. In some embodiments, this document also provides compositions, articles, and kits comprising cell populations, the cell populations comprising cell populations classified as having a desired differentiation state by any of the provided methods. In some embodiments, this document also provides a method for implanting a cell population having a desired differentiation state (e.g., classified according to any of the provided methods) into a subject.

[0077] All publications mentioned in this application, including patent documents, scientific papers, and databases, are incorporated herein by reference in their entirety for all purposes, as if each individual publication were incorporated individually by reference. If any definition presented herein contradicts or otherwise is inconsistent with a definition presented in a patent, application, published application, or other publication incorporated herein by reference, the definition presented herein shall prevail over the definition incorporated herein by reference.

[0078] The chapter titles used in this article are for organizational purposes only and should not be construed as limiting the topics described.

[0079] definition Unless otherwise defined, all specialized terms, symbols, and other technical and scientific terms or terminology used herein are intended to have the same meaning as commonly understood by one of ordinary skill in the art to which the claimed subject matter pertains. In some instances, for clarity and / or convenience of reference, terms with commonly understood meanings are defined herein, and the inclusion of these definitions herein should not be construed as indicating a material difference from those commonly understood in the art.

[0080] As used herein, unless the context clearly indicates otherwise, the singular forms “a,” “an,” and “the” include plural indicators. For example, “a” or “an” means “at least one” or “one or more.” It should be understood that aspects and variations described herein include “consisting of aspects and variations” and / or “substantially consisting of aspects and variations.”

[0081] Throughout this disclosure, all aspects of the claimed subject matter are presented in a scope format. It should be understood that this scope format is for convenience and brevity only and should not be construed as a rigid limitation on the scope of the claimed subject matter. Therefore, the scope description should be considered as having specifically disclosed all possible sub-scopes and individual values ​​within that scope. For example, in the case of providing a range of values, it should be understood that every intermediate value between the upper and lower limits of the range, as well as any other specified value or intermediate value within that specified range, is covered within the claimed subject matter. The upper and lower limits of these smaller scopes may be independently included within the smaller scopes and are also covered within the claimed subject matter, conditional on any explicitly excluded limit value within the specified scope. When a specified scope includes one or two limits, the scope excluding any one or both of those included limits is also included in the claimed subject matter. This applies regardless of the width of the scope.

[0082] As used herein, the term "about" refers to the typical range of error for individual values ​​that are readily known. References to "about" values ​​or parameters include (and describe) implementations for that value or parameter itself. For example, a description of "about X" includes a description of "X".

[0083] As used herein, the statement that a cell or cell population “expresses” a particular marker or is “positive” for a particular marker means that the particular marker is present on or detectably within the cell. When referring to a surface marker, the term means the presence of surface expression as detected by flow cytometry, for example by staining with an antibody that specifically binds to the marker and detecting said antibody, wherein the staining is detectable by flow cytometry and the level is substantially higher than that detected under otherwise identical conditions with a type-matched control using the same procedure, and / or the level is substantially similar to that of cells known to be positive for the marker, and / or the level is substantially higher than that of cells known to be negative for the marker. When referring to a marker within a cell, such as a transcription product or translation product, the term means, for example, the presence of a detectable transcription product or translation product, wherein the level of the detected product is substantially higher than that detected under otherwise identical conditions with a control using the same procedure, and / or the level is substantially similar to that of cells known to be positive for the marker, and / or the level is substantially higher than that of cells known to be negative for the marker.

[0084] As used herein, the statement that a cell or cell population “does not express” a particular marker or is “negative” for a particular marker means that the particular marker is substantially undetectable on or within the cell. When referring to a surface marker, the term means the absence of surface expression as detected by flow cytometry, for example by staining with an antibody that specifically binds to the marker and detecting said antibody, wherein the staining is not detected by flow cytometry at a level substantially higher than that detected by the same procedure with a type-matched control under otherwise identical conditions, and / or at a level substantially lower than that detected by cells known to be positive for the marker, and / or at a level substantially similar to that detected by cells known to be negative for the marker. When referring to markers within cells, such as transcripts or translation products, the term means, for example, the absence of detectable transcripts or translation products, wherein the product is not detected at a level substantially higher than that detected by the same procedure with a control under otherwise identical conditions, and / or at a level substantially lower than that detected by cells known to be positive for the marker, and / or at a level substantially similar to that detected by cells known to be negative for the marker.

[0085] As used in this article when referring to genes, the terms "expressed" or "expressed" refer to the transcriptional and / or translational products of that gene. The expression level of a DNA molecule in a cell can be determined based on the amount of corresponding mRNA present in the cell or the amount of protein encoded by that DNA produced by the cell. RNA sequencing (RNA-seq) is commonly used to determine gene expression levels. See, for example, Conesa et al. (2016). Genome BiologyA review of RNAseq methods in 17:13 (https: / / doi.org / 10.1186 / s13059-016-0881-8).

[0086] As used herein, the term "stem cell" refers to a cell characterized by its ability to self-renew through mitosis and its potential to differentiate into any of a variety of cell types. In mammalian stem cells, embryonic stem cells and somatic stem cells can be distinguished. Embryonic stem cells reside in the blastocyst and produce embryonic tissue, while somatic stem cells reside in adult tissue and are used for tissue regeneration and repair.

[0087] "Self-renewal" refers to the ability of a cell to divide and produce at least one daughter cell with the self-renewal properties of its parent cell. This second daughter cell can then enter a specific differentiation pathway. For example, a self-renewing hematopoietic stem cell can divide and form one daughter stem cell and another daughter cell, which then differentiates into its final form in the bone marrow or lymphatic pathway.

[0088] As used herein, the term "progenitor cell" refers to a cell that has the potential to differentiate into any of a variety of cell types but has lost the ability to self-renew relative to stem cells. For example, a progenitor cell can produce two daughter cells after cell division, which exhibit a more differentiated (e.g., restricted) phenotype.

[0089] As used in this article, the term "non-self-renewing cell" refers to a cell that undergoes cell division to produce daughter cells, none of which have the differentiation potential of the parent cell type, such as producing differentiated daughter cells.

[0090] As used herein, the term "adult stem cell" refers to undifferentiated cells found in an individual after embryonic development. Adult stem cells multiply through cell division to replenish dead cells and regenerate damaged tissue. Adult stem cells are capable of dividing and producing another cell identical to themselves or producing more differentiated cells. Although adult stem cells are associated with the expression of pluripotency markers such as Rex1, Nanog, Oct4, or Sox2, they do not possess the ability of pluripotent stem cells to differentiate into cell types from all three germ layers.

[0091] As used herein, the term "pluripotency" or "pluripotency" refers to cells capable of producing progeny that, under appropriate conditions, differentiate into cell types that collectively exhibit characteristics associated with cell lineages from the three germ layers (endoderm, mesoderm, and ectoderm). Pluripotent stem cells can contribute to the tissue organization of prenatal, postnatal, or adult organisms.

[0092] As used herein, the term "pluripotent stem cell characteristic" refers to the cellular characteristics that distinguish pluripotent stem cells from other cells. The expression or non-expression of certain combinations of molecular markers is an example of pluripotent stem cell characteristics. More specifically, human pluripotent stem cells may express at least some, optionally all, of the following non-limiting list of markers: SSEA-3, SSEA-4, TRA-1-60, TRA-1-81, TRA-2-49 / 6E, ALP, Sox2, E-cadherin, UTF-1, Oct4, Lin28, Rex1, and Nanog. Cell morphology associated with pluripotent stem cells is also a characteristic of pluripotent stem cells.

[0093] As used herein, the terms “induced pluripotent stem cells,” “iPS,” and “iPSC” refer to pluripotent stem cells artificially obtained (e.g., through artificial manipulation) from non-pluripotent cells. “Non-pluripotent cells” can be cells with a lower self-renewal and differentiation potential than pluripotent stem cells. Cells with lower potential can be adult stem cells, tissue-specific progenitor cells, primary cells, or secondary cells.

[0094] As used herein, the terms "specification" or "specified" refer to the narrowing down of cell or tissue fate to a finite number of specific cell types. Specified cells can still alter their specific fate until they reach a oriented state. Specified cells are capable of autonomous differentiation (e.g., self-differentiation) when placed in an environment relatively neutral to the developmental pathway, such as a petri dish or test tube. Cell morphology can still be altered during the specification phase. If a specified cell is transplanted into a population of cells specified in a different manner, its interaction with the new neighboring cells can alter the fate of the transplanted cell.

[0095] As used herein, "oriented state" refers to a cell possessing only one type of cell it can differentiate into. For example, oriented dopaminergic cells cannot become other types of neurons, although they may not yet be dopaminergic neurons themselves, and may or may not express specific markers of dopaminergic neurons. Oriented cells may also be capable of autonomous differentiation when placed in embryonic regions unrelated to the cells themselves. For example, unrelated regions for oriented dopaminergic cells are any organ or tissue other than the brain. Oriented cells can also be capable of autonomous differentiation when placed in clusters of cells that are specialized in different ways in a culture dish.

[0096] As used in this article, the terms “differentiated” or “classified” refer to one or more cells that have acquired cell type-specific functions.

[0097] "Neuron precursor cells" are cells that have the tendency to differentiate into neurons or glial cells but do not possess the pluripotent potential of stem cells. Neuron precursors are cells that are characterized as neurons or glial lineages, characterized by expressing one or more marker genes specific to the neuron or glial lineage. The terms "neural" and "neuronal" are used according to their common meaning in the art and are used interchangeably throughout the text.

[0098] As used herein, "dopaminergic cell" or "differentiated dopaminergic cell" refers to a cell capable of synthesizing the neurotransmitter dopamine. In this embodiment, the dopaminergic cell is an A9 dopaminergic cell. The term "A9 dopaminergic cell" refers to the most densely packed group of dopaminergic cells in the human brain, located in the substantia nigra pars compacta of the midbrain of a healthy adult.

[0099] As used herein, the term "directed dopaminergic cell" refers to a cell that differentiates into dopaminergic neurons but not into non-dopaminergic cells. A "directed dopaminergic cell" is a cell capable of differentiating into dopaminergic neurons independently of its environment. Directed dopaminergic cells may express Foxa2 or Nurl1. Directed dopaminergic cells may not express serotonin.

[0100] As used in this article, the term "reprogramming" refers to the process of dedifferentiating non-pluripotent cells into cells that exhibit pluripotent stem cell characteristics.

[0101] As used herein, the term "cell culture" can refer to a population of cells outside an organism. Cell cultures can be established from primary cells isolated from a cell bank or animal, or from secondary cells derived from one of these sources and immortalized to obtain long-term in vitro cultures.

[0102] As used herein, the terms “culture,” “culturing,” “grow,” “growing,” “maintain,” “maintaining,” “expand,” and “expanding” are used interchangeably when referring to cell cultures themselves or the culturing process, meaning cells maintained in vitro (e.g., in vitro) under conditions suitable for survival. This means that the cultured cells are allowed to survive, and that the culture can induce cell growth, differentiation, or division.

[0103] As used herein, the term "adherent culture dish" refers to a culture dish in which cells can attach to the dish via extracellular matrix molecules, and where separating the cells from the dish requires the use of enzymes (e.g., trypsin, dispersant, etc.). "Adherent culture dishes" are quite different from such dishes; cell attachment is reduced, and removing the cells from the dish does not require the use of enzymes.

[0104] As used herein, the term "non-adherent culture vessel" refers to a culture vessel in which cell adhesion is reduced or restricted (e.g., for a period of time). Non-adherent culture vessels may contain low- or ultra-low-adherence surfaces, such as those achieved by treating the surfaces with a substance that prevents cell adhesion, such as a hydrogel (e.g., a neutrally charged hydrogel and / or a hydrophilic hydrogel) and / or a surfactant (e.g., prawnic acid). Non-adherent culture vessels may contain round or concave pores, and / or micropores (e.g., Aggrewells). ™ In some implementations, non-adherent culture dishes are Aggrewell. ™ For non-adherent culture dishes, it may not be necessary to use enzymes to remove cells from the culture dish.

[0105] As used herein, a composition refers to any mixture of two or more products, substances, or compounds, including cells. It can be a solution, suspension, liquid, powder, paste, aqueous solution, non-aqueous solution, or any combination thereof.

[0106] The term "pharmaceutical composition" refers to a composition suitable for pharmaceutical use, such as in mammalian subjects (e.g., humans). Pharmaceutical compositions typically contain an effective amount of an active agent (e.g., cells) and a carrier, excipient, or diluent. This carrier, excipient, or diluent is typically pharmaceutically acceptable, respectively.

[0107] "Pharmaceutically acceptable carriers" refer to components in a drug formulation that are non-toxic to the subjects, excluding the active ingredient. Pharmaceutically acceptable carriers include, but are not limited to, buffers, excipients, stabilizers, or preservatives.

[0108] The term "instructions for use" is used to refer to instructions that are typically included in the commercial packaging of a therapeutic product, which contain information about the indications, usage, dosage, administration, combination therapy, contraindications and / or warnings for the use of such therapeutic products.

[0109] As used in this article, “subject” is a mammal, such as a human or other animal, and is usually a human.

[0110] Methods for obtaining neuronal progenitor cells and determining gene expression levels In some embodiments, this document provides methods for predicting one or more properties of an in vitro population of neuronal progenitor cells. In some embodiments, the provided methods are used to identify or select in vitro populations of neuronal progenitor cells that have or are capable of producing cells with one or more of these properties. For example, in some embodiments, the invention provides methods for predicting the successful transplantation and survival of neuronal progenitor cells after implantation into the brain of a subject. In some embodiments, the invention provides methods for predicting whether neurons derived from a population of neuronal progenitor cells will produce dopamine.

[0111] pluripotent stem cells In some embodiments, the neuronal progenitor cell population is obtained by differentiating pluripotent stem cells. In some embodiments, the pluripotent stem cells are embryonic stem (ES) cells, induced pluripotent stem cells (iPSCs), or a combination thereof. In some embodiments, the pluripotent stem cells are human induced pluripotent stem cells. In some embodiments, the pluripotent stem cells are autologous to the subject. In some embodiments, the pluripotent stem cells are allogeneic to the subject. In some embodiments, the pluripotent stem cells are derived from healthy human subjects. In some embodiments, the pluripotent stem cells are derived from human subjects suffering from a neurodegenerative disease or condition. In some embodiments, the neurodegenerative disease or condition includes the loss of dopaminergic neurons. In some embodiments, the neurodegenerative disease or condition is Parkinson's syndrome. In some embodiments, the neurodegenerative disease or condition is Parkinson's disease.

[0112] Methods for generating iPSCs are known. For example, iPSCs can be generated through a process called reprogramming, in which non-pluripotent cells are effectively “dedifferentiated” into an embryonic stem cell-like state by engineering them to express genes such as OCT4, SOX2, and KLF4. (Takahashi and Yamanaka (2006)) Cell 126: 663-76. In some implementations, fibroblasts can be reprogrammed into iPSCs by converting them with genes (OCT4, SOX2, NANOG, LIN28, and KLF4) cloned into plasmids (e.g., see Yu et al., (2007)). Science 318: 1917-1920). In some implementations, cells are reprogrammed using a non-integrating Sendai virus (e.g., using CTS). ™ CytoTune ™The iPS2.1 Sendai Reprogramming Kit reprograms non-pluripotent cells, such as fibroblasts, to induced pluripotent stem cells. In some embodiments, the pluripotent stem cells are iPSCs artificially derived from the subject's non-pluripotent cells. In some embodiments, the non-pluripotent cells are fibroblasts.

[0113] Methods for differentiating neuronal cells In some embodiments, a population of neuronal progenitor cells is differentiated from pluripotent stem cells under conditions conducive to cellular neural differentiation. Suitable differentiation methods for obtaining neuronal progenitor cells are known to those skilled in the art. Such methods may involve direct (e.g., by delivering a gene payload into the cells) or indirect (e.g., by using a variety of pharmacological agents to tilt the differentiation pathway toward neuronal fate (Telias (2023)). Neural Regen.ResManipulating gene expression. The dual SMAD inhibition protocol is an example of the latter approach to neurally differentiating pluripotent stem cells into neuronal progenitor cells (see, for example, U.S. Patent Publication US2019 / 0211306). This process may involve exposing pluripotent stem cells to: (a) a bone morphogenetic protein (BMP) signaling inhibitor; (b) a TGF-β / activin Nodal signaling inhibitor; and (c) at least one sound hedgehog factor (SHH) signaling activator. The method may also include exposing pluripotent stem cells to at least one GSK3β signaling inhibitor. Methods for neural differentiation of pluripotent stem cells are also described in, for example, the following patents: U.S. Patent Publication US2023 / 0059010 entitled “METHODS OF DIFFERENTIATING NEURAL CELLS AND RELATED COMPOSITIONS AND METHODS OF USE”; U.S. Patent Publication US2024 / 0329032 entitled “METHODS OF DIFFERENTIATING NEURAL CELLS AND PREDICTING ENGRAFTMENT THEREOF AND RELATED COMPOSITIONS”; and U.S. Patent Application 18 / 742,917 entitled “METHODS FOR DIFFERENTIATING DOPAMINERGIC NEURONS FROM STEM CELLS”, filed June 13, 2024. Additional exemplary methods for differentiating stem cells into in vitro neuronal cells are described, for example, in WO2014 / 176606, US8460931, US10273453, WO2012 / 095730, US9309495, US2019 / 0249140, US2018 / 0298326, WO2009 / 148170, WO2021 / 146349, WO2021 / 216623, WO2021 / 216622, WO2013 / 104752, WO2010 / 096496, WO2013 / 067362, WO2016 / 196661, WO2015 / 143342 and US2016 / 0348070.

[0114] Direct genetic manipulation of stem cells to induce neuronal differentiation can involve the introduction of nucleic acids encoding neuronal genes, and / or regulatory sequences such as promoters and enhancers. Such methods can involve introducing nucleic acid constructs into pluripotent stem cells via (i) plasmid transfection, lipid transfection, or electroporation, or (ii) the use of viral vectors, such as adeno-associated virus (AAV) or lentiviral vectors. Lentiviral-based neuronal differentiation methods are described, for example, in the following literature: Zhang et al., (2013). Neuron 78: 785-798.

[0115] In some embodiments, differentiation conditions are achieved through culture including adherent cell culture, such as using methods described, for example, in U.S. Patent Publication US2019 / 0211306. In these embodiments, neuronal progenitor cells are sometimes obtained between day 18 and day 24 after the start of the differentiation process. In some embodiments, differentiation conditions include suspension cell culture, as described herein and in U.S. Patent Publication 2023 / 0059010 and U.S. Patent Application 18 / 742,917. In these embodiments, neuronal progenitor cells are sometimes obtained between day 13 and day 20 after the start of the differentiation process (such as day 16 or about day 16). In some embodiments, the pluripotent stem cells are induced pluripotent stem cells (iPSCs). In some embodiments, the pluripotent stem cells are autologous to the subject to whom neuronal progenitor cells are to be implanted.

[0116] Determine gene expression levels In some embodiments, gene expression levels in cells of any test sample or reference cell population described herein are determined based on the level of a gene product synthesized using information encoded by one or more genes. In some embodiments, the gene product is any biomolecule assembled, generated, and / or synthesized using gene-encoded information, and may include polynucleotides and / or peptides. In some embodiments, assessing, measuring, and / or determining gene expression includes determining or measuring the level, amount, or concentration of the gene product. In some embodiments, the level, amount, or concentration of the gene product may be transformed (e.g., normalized) or directly analyzed (e.g., raw).

[0117] In some embodiments, the gene product includes a protein encoded and / or expressed by a gene, i.e., a polypeptide. In certain embodiments, the gene product encodes a protein that is confined and / or exposed on the cell surface. In some embodiments, the protein is a soluble protein. In some embodiments, the protein is secreted by the cell. In certain embodiments, gene expression is the amount, level, and / or concentration of the protein encoded by the gene. In some embodiments, one or more protein gene products are measured by any suitable method. Suitable methods for assessing, measuring, determining, and / or quantifying the level, amount, or concentration of multiple or more protein gene products include detection using immunoassays, nucleic acid-based or protein-based aptamer techniques, HPLC (high-precision liquid chromatography), peptide sequencing (such as Edman degradation sequencing) or mass spectrometry (such as MS / MS, optionally coupled to HPLC), and any of the aforementioned microarray adaptations (including nucleic acid, antibody, or protein-protein (i.e., non-antibody) arrays). In some embodiments, the immunoassay is or includes a method or assay based on an immune response, such as detecting a protein by detecting the binding of an antibody or antigen-binding antibody fragment to the gene product. Immunoassays include quantitative immunocytochemistry or immunohistochemistry, ELISA (including direct, indirect, sandwich, competitive, multiplex, and portable ELISA (see, for example, U.S. Patent No. 7,510,687), Western blotting (including one-dimensional, two-dimensional, or high-dimensional blotting or other chromatographic methods, optionally including peptide sequencing), enzyme immunoassay (EIA), RIA (radioimmunoassay), and SPR (surface plasmon resonance).

[0118] In some embodiments, the gene product is a polynucleotide encoded by a gene, such as mRNA or a protein. In some embodiments, the gene product is a polynucleotide expressed and / or encoded by a gene. In some embodiments, the polynucleotide is RNA. In some embodiments, the gene product is messenger RNA (mRNA), transfer RNA (tRNA), ribosomal RNA, small nuclear RNA, small nucleolar RNA, antisense RNA, long noncoding RNA, microRNA, Piwi-interacting RNA, small interfering RNA, and / or short hairpin RNA. In a particular embodiment, the gene product is mRNA.

[0119] In certain embodiments, assessing, measuring, determining, and / or quantifying the amount or level of RNA gene products includes steps of generating, polymerizing, and / or derivatizing cDNA polynucleotides and / or cDNA oligonucleotides from the RNA gene product. In some embodiments, the RNA gene product is assessed, measured, determined, and / or quantified by directly assessing, measuring, determining, and / or quantifying cDNA polynucleotides and / or cDNA oligonucleotides derived from the RNA gene product.

[0120] In certain embodiments, the amount or level of polynucleotides in a sample can be assessed, measured, determined, and / or quantified by any suitable method. For example, in some embodiments, the amount or level of polynucleotide gene products can be assessed, measured, determined, and / or quantified by: polymerase chain reaction (PCR), including reverse transcriptase (rt) PCR, droplet digital PCR, real-time and quantitative PCR (qPCR) methods (including, for example, TAQMAN). ® Molecular beacons, Lightup ™ SCORPION ™ SIMPLEPROBES ® See, for example, U.S. Patents 5,538,848; 5,925,517; 6,174,670; 6,329,144; 6,326,145 and 6,635,427); northern blotting; for example, Southern blotting of reverse transcripts and their derivatives; array-based methods, including blotting arrays, microarrays or in situ synthetic arrays; and sequencing, such as sequencing-by-synthesis, pyrosequencing, dideoxy sequencing or ligation-by-sequencing, or other methods (such as those discussed in Shendure et al., (2004)). Nat. Rev. Genet .5:335-44 or Nowrousian (2010) Euk.Cell 9(9): 1300-1310), including specific platforms such as HELICOS ® ROCHE ® 454, ILLUMINA ® / SOLEXA ® ABI SOLiD ® and POLONATOR ® Sequencing. In a particular implementation, the level of nucleic acid gene products is measured using quantitative PCR (qPCR) methods, such as qRT-PCR.

[0121] In certain embodiments, the method used to determine gene expression is a quantitative method. In some embodiments, the method provides relative gene expression levels. In some embodiments, the method used to measure the relative amount of mRNA expression is reverse transcription quantitative polymerase chain reaction (RT-PCR, followed by qPCR). RT-PCR initially generates a complementary DNA (cDNA) template from mRNA via reverse transcription. The cDNA template is then used for qPCR, where the fluorescence of the probe changes as DNA amplification proceeds. Using a standard curve, qPCR enables the quantification of the relative levels of mRNA species within a sample. RT-qPCR assays employ fluorescent reporter probes (i.e., TaqMan, Life Technologies), which can be designed for specific mRNA targets, resulting in minimal cross-reactivity and high specificity.

[0122] In some embodiments, total RNA is isolated using standard protocols, such as the RNeasy Mini Kit (Qiagen Inc.) with a gDNA removal column. The total mRNA from each sample is converted to complementary DNA (cDNA) using available methods, such as the High Capacity RNA to cDNA Kit (Life Technologies). Based on an estimated number of cells collected, the cDNA may be pre-amplified, such as using the TaqManPreAmp Master Mix Kit (Invitrogen) prior to qPCR, to amplify the cDNA targets equally without introducing bias, while increasing the amount of total cDNA that may be needed when measuring multiple targets. In some embodiments, the expression levels of the G genes TTR, PRR16, and CD47 are assessed using the TaqMan Gene Expression Assay Kit (Life Technologies) which is specific for these target proteins. In some embodiments, the expression levels of the D genes B3GALT5, GREM2, and FRMPD3 are assessed using the TaqMan Gene Expression Assay Kit (Life Technologies) which is specific for these target proteins.

[0123] In some implementations, target gene expression can be expressed in multiple ways, including as a total score of target gene expression relative to control gene expression or as expression of a single target gene. In some implementations, RT-qPCR data are expressed as “cycles to threshold” (Ct). As understood by those skilled in the art, Ct is a relative value representing the number of cycles at which the fluorescence signal of the amplified DNA reaches a defined threshold level exceeding the background. Due to variability between assays and differences in the number of cells from which mRNA is extracted, Ct values ​​are typically normalized to Ct amplification values ​​for constitutive expression reference sequences (such as housekeeping genes). Therefore, differential expression is considered on a gene-by-gene basis and is expressed as a normalized Ct value (ACt) of biological repeatability between sample groups. As described herein, the normalized Ct value is also referred to as AACt. In the resulting expression, the Ct level is inversely proportional to the amount of target nucleic acid in the sample, such that a high ACt value indicates low expression of a given gene, while a highly expressed gene has a low ACt value.

[0124] In some embodiments, gene expression is relative gene expression. In some embodiments, the relative gene expression of a target gene is determined as the ratio of the corresponding target gene (e.g., the G gene or the D gene) to a reference gene. In some embodiments, the reference gene is a housekeeping gene. In some embodiments, the reference gene is selected from PRS18, IPO8, RPL113A, HSP90AB1, UBC, PSMC4, SDHA, HPRT1, HMBS, TFRC, PPIA, RPL30, GUSB, ACTB, LDHA, RPS17, GAPDH, PPIH, NONO, PUM1, HBB, G6PD, TBP, ALAS1, PGK1, CDKN1A, YWHAZ, POP4, RPLP0, or B2M. In some embodiments, the reference gene is a housekeeping gene, which is GAPDH.

[0125] mRNA levels can also be quantified using several other methods, including Northern blotting, which provides information about the size and sequence of mRNA molecules and includes methods to distinguish transcripts from alternative splicing. Other methods known in the art include the use of DNA microarrays and techniques such as sequenced gene expression analysis (SAGE) to provide relative measurements of different mRNAs.

[0126] In a particular implementation, the expression of two or more genes is measured or evaluated simultaneously.

[0127] In some embodiments, multiplex PCR, such as multiplex rt-PCR assessment or multiplex quantitative PCR (qPCR), is used to measure, determine, and / or quantify the level, amount, or concentration of two or more gene products. In some embodiments, microarrays (e.g., AFFYMETRIX) are used.® AGILENT ® and ILLUMINA ® qRT-PCR (qRT-PCR array) is used to assess, measure, determine, and / or quantify the level, amount, or concentration of two or more gene products. In some implementations, qRT-PCR uses a set of three nucleic acids for each gene, wherein the three nucleic acids include a primer pair and a probe that binds between the target nucleic acid regions to which the primers bind (commercially known as TAQMAN). ® (Determination method).

[0128] In some implementations, the microarray is used to assess, measure, determine, and / or quantify the level, amount, or concentration of cDNA polynucleotides derived from RNA gene products.

[0129] In some embodiments, the expression of one or more gene products (e.g., polynucleotide gene products) is determined by sequencing the gene product and / or by sequencing the cDNA polynucleotides derived from the gene product. In some embodiments, sequencing is performed using non-Sanger sequencing methods and / or next-generation sequencing (NGS) technologies. Examples of next-generation sequencing technologies include massively parallel signature sequencing (MPSS), Polony sequencing, pyrosequencing, reversible dye terminator sequencing, SOLiD sequencing, ion semiconductor sequencing, DNA nanosphere sequencing, Helioscope single-molecule sequencing, single-molecule real-time (SMRT) sequencing, single-molecule real-time (RNAP) sequencing, and nanopore DNA sequencing.

[0130] In some implementations, NGS technology is RNA sequencing (RNA-Seq). In specific implementations, the expression of one or more polynucleotide gene products is measured, determined, and / or quantified by RNA-Seq. RNA-Seq, also known as whole transcriptome shotgun sequencing, determines the presence and quantity of RNA in a sample. RNA sequencing methods are compatible with the most common DNA sequencing platforms [HiSeq system (Illumina), FLX 454 genome sequencer system (Roche), Applied Biosystems SOLiD (Life Technologies), IonTorrent (Life Technologies). These platforms require RNA to be initially reverse transcribed into cDNA. Conversely, the single-molecule sequencer HeliScope (Helicos BioSciences) can use RNA as a sequencing template. Validation of the principle for direct RNA sequencing on the PacBio RS platform has also been demonstrated (Pacific Bioscience). In some implementations, one or more RNA gene products are evaluated, measured, determined, and / or quantified by RNA-seq. In some implementations, RNA-seq is tag-based RNA-seq. In tag-based methods, each transcript is represented by a unique tag. Initially, tag-based methods were developed as sequence-based approaches to measure transcript abundance and identify differentially expressed genes, assuming that the number (count) of tags directly corresponds to the abundance of mRNA molecules. The reduced complexity of samples obtained by sequencing defined regions is crucial for making Sanger-based methods affordable. When NGS technology became available, the large number of reads that could be generated facilitated the analysis of differentially expressed genes. Tagged-based methods do not encounter transcript length biases, such as those observed in shotgun sequencing for quantifying gene expression levels. By definition, all tag-based methods are strand-specific. In specific implementations, tag-based RNA-seq is used to assess, measure, identify, and / or quantify one or more RNA gene products.

[0131] In some implementations, RNA-seq is shotgun RNA-seq. Many protocols have been described for shotgun RNA-seq, but they share several common steps: fragmentation (which can occur at the RNA or cDNA level), RNA to cDNA conversion (via oligodT or random primers), second-strand synthesis, ligation of adaptor sequences at the 3' and 5' ends (at the RNA or DNA level), and final amplification. In some implementations, if poly(A)+ RNA is selected prior to fragmentation, RNA-seq can focus solely on polyadenylated RNA molecules (primarily mRNA, but also some lncRNA, snoRNA, pseudogenes, and histones), or, if no selection is made, it can include non-polyadenylated RNA. In the latter case, ribosomal RNA (over 80% of the total RNA library) needs to be depleted before fragmentation. Therefore, it is evident that differences in capturing transcriptome mRNA portions lead to partial overlap in the types of transcripts detected. Moreover, different protocols can affect the abundance and distribution of sequencing reads. This makes it difficult to compare experimental results performed with different library preparation protocols.

[0132] In some implementations, RNA is obtained from each sample, fragmented, and used to generate complementary DNA (cDNA) samples, such as cDNA libraries for sequencing. Reads can be processed and aligned with the human genome, and the expected number of mappings for each gene / isotype can be estimated and used to determine read counts. In some implementations, read counts are normalized according to gene / isotype length and the number of reads in the library to obtain, for example, FPKM normalized according to gene / isotype length and the number of reads in the library, to obtain the number of fragments per thousand bases of exons per million mapped reads (FPKM) based on gene length and total mapped reads. In some aspects, inter-sample normalization is achieved through normalization, such as 75th quantile normalization, where each sample is scaled by the median of the 75th quantile from all samples, for example, to obtain a quantile-normalized FPKM (FPKQ) value. A logarithmic transformation (log2) can be applied to the FPKQ values.

[0133] In some embodiments, RNA is obtained from each sample, fragmented, and used to generate complementary DNA (cDNA) samples, such as cDNA libraries for sequencing. Reads can be processed and aligned with the human genome, and the expected number of mappings for each gene / isotype can be estimated and used to determine read counts. In some embodiments, read counts are normalized according to gene / isotype length and the number of reads in the library. In some embodiments, read counts are provided in the form of counts per million (CPM). In some embodiments, the CPM read counts are logarithmically transformed (e.g., log2).

[0134] In some embodiments, relative gene expression is measured by comparing the CPM of the target gene to the CPM of a reference gene (such as a housekeeping gene). In some embodiments, the housekeeping gene is GAPDH. In some embodiments, the relative gene expression of the target gene is determined as the ratio of the CPM of the target gene to the CPM of the housekeeping gene (e.g., GAPDH).

[0135] In some embodiments, microarray analysis is used to obtain gene expression levels. In some embodiments, RNA sequencing is used to obtain gene expression levels. In some embodiments, both microarray analysis and RNA sequencing are used to obtain gene expression levels. In some embodiments, RNA sequencing is performed on bulk RNA from multiple cells. In some embodiments, bulk RNA sequencing data is obtained from merged RNA from multiple cells. In some embodiments, RNA sequencing is performed on single cells. In some embodiments, RNA sequencing is performed on both bulk RNA from multiple cells and single cells.

[0136] Any suitable method for obtaining bulk RNA sequencing data can be used (e.g., see Chao et al., (2019)). BMC Genomics (20: 571). For example, total RNA from a sample (e.g., multiple cells from a cell population) can be isolated using TRIZOL, treated with DNase I, and purified. The concentration and quality of the isolated RNA can be measured and checked before library preparation of the total RNA or mRNA. For library preparation, the total RNA or mRNA can be fragmented and converted to cDNA using reverse transcription. After double-stranded cDNA construction, amplification, and optional barcoding, the library can be processed for next-generation sequencing using any suitable library preparation technology, sequencing platform, and genome alignment tool.

[0137] In some embodiments, single-cell RNA sequencing is used to obtain gene expression levels. In some embodiments, the use of single-cell RNA sequencing data provides certain advantages. In some embodiments, the use of single-cell RNA sequencing data allows for the characterization of cell subpopulations, such as directional dopaminergic cells within a larger cell population. In some embodiments, the use of single-cell RNA sequencing data reduces the number of cells required in the methods provided herein, for example, reducing the number of cells required to obtain data for training machine learning models. In some embodiments, the use of single-cell RNA sequencing data improves the characterization of biological variability across cells. In some embodiments, the use of single-cell RNA sequencing data allows for easier validation and interpretation of gene expression levels.

[0138] Any suitable method for single-cell RNA sequencing can be used (e.g., see Zheng et al., (2017)). Nature Communications 8: 14049, and Haque et al., (2017) Genome Medicine 9:75). For single RNA sequencing, single cells can be isolated from a sample (e.g., an in vitro cell population) using flow cytometry cell sorting, microfluidic platforms, or droplet-based methods. The isolated cells are lysed to allow capture of RNA molecules. Poly[T] primers can be specifically used to analyze polyadenylated mRNA molecules, and reverse transcription is used to convert the induced mRNA molecules into cDNA. In some cases, unique molecular identifiers can be used to label individual mRNA molecules based on cell origin. The cDNA library can then be amplified, optionally barcoded, and sequenced, for example using next-generation sequencing (NGS) and library preparation technologies, sequencing platforms, and genome alignment tools (similar to those used for bulk RNA samples). In some cases, unbiased cell type classification within mixed populations of different cell types can be achieved with as few as 10,000 to 50,000 reads / cell, and single-cell libraries from various common protocols can approach saturation at sequencing depths of 1,000,000 reads.

[0139] In some embodiments, gene expression levels include bulk RNA sequencing data and single-cell RNA sequencing data. In some embodiments, the bulk RNA sequencing data and single-cell RNA sequencing data are obtained from the same cell population. In some embodiments, single-cell RNA sequencing data can be used to estimate bulk RNA sequencing data obtained from the same cell population. In some embodiments, the estimated bulk RNA sequencing data is obtained by averaging single-cell RNA sequencing data from cells within the same cell population. In some embodiments, gene expression levels include the estimated bulk RNA sequencing data.

[0140] Methods for determining the characteristics of neuronal progenitor cells In some embodiments, the provided method includes using gene expression levels of multiple genes in one or more cells of a neuronal progenitor cell population to predict whether the neuronal progenitor cell population will survive transplantation after implantation into the brain of a subject. In other embodiments, the provided method includes using gene expression levels of multiple genes in one or more cells of a neuronal progenitor cell population to predict whether the neuronal progenitor cell population will produce dopamine-producing neurons after implantation into the brain of a subject.

[0141] The methods provided are based on the finding that the expression levels of certain genes in a population of neuronal progenitor cells are correlated with the predicted characteristics of cells derived from these cells after cell implantation in a subject. These methods mitigate the variability that may exist when generating cell products differentiated from pluripotent stem cells. Consistency of administered cell products is important for maximizing the efficacy of cell products in multiple different subjects. This is especially true for autologous cell therapies, where there can be higher variability due to donor differences and a degree of variability that may exist during manufacturing. The methods provided produce more consistent cell products with higher confidence, thereby improving treatment options for subjects, particularly those receiving neuronal progenitor cells derived from their own cells. The ability to assess whether progenitor cells will generate neurons with desired characteristics after implantation based on their gene expression levels prior to implantation for therapeutic purposes is a significant advance in the success of pluripotent stem cell-derived cell therapies.

[0142] Predicting transplant survival In some embodiments, the present invention provides a method for predicting whether a population of neuronal progenitor cells is likely to successfully transplant and survive in a brain region after implantation in a subject. In some embodiments, the prediction indicates the ability of cells to transplant and survive after implantation and form mature dopamine neuron grafts of a certain size. In some embodiments, the gene expression level of one of a plurality of genes in a test sample of the neuronal progenitor cell population is used to predict cell transplantation survival of cells derived from neuronal progenitor cells, wherein the gene is associated with survival ability after implantation in a subject (hereinafter referred to as the “G gene”). Exemplary G genes are described herein and are included in Table E1. In some embodiments, the G gene is selected from the group consisting of: AC000120.3, KRT77, TTR, PRR16, MEGF10, PDE3A, GDPD2, CMTM8, APOA1, CMTM7, CDHR3, CORIN, VTN, CPNE8, EFEMP1, CD47, SPARC, JAM2, CDO1, PLXDC2, DYNLL2, ITGA3, RPS6KL1, CHRNB2, S ULT4A1, PTPN3, LZTS1, RUNX1T1, TMEM145, EPHA10, CARMIL3, MANEAL, TMEM176B, MPP3, DRAXIN, ADGRB1, K IF26A, CELF5, CNTN2, ASPHD1, SVOP, ANGPT2, SLC22A15, SRRM3, GRIN2D, DACH2, CHST1, GRIN1, LHX5 and NOS2.

[0143] In some embodiments, these methods include predicting the neuronal transplant viability of neuronal progenitor cells by associating determined gene expression levels of one or more G genes in a test sample with a reference map of each G gene, the reference map relating graft size to gene expression levels of G genes in a training set including one or more reference samples. In some embodiments, the gene expression levels of G genes are associated with the ability of a population of neuronal progenitor cells to produce cells that transplant and survive in a brain region after the population of neuronal progenitor cells is implanted into a brain region of a subject. In some embodiments, a population of neuronal progenitor cells for implantation for therapeutic purposes is selected based on the gene expression levels of one or more G genes.

[0144] In some implementations, each data point on the reference plot is determined by the following steps: (a) measuring the gene expression level of the G gene in a reference sample comprising a population of neuronal progenitor cells; (b) implanting neuronal progenitor cells from the reference sample into a brain region of a test animal and measuring the size of the graft formed by the implanted neuronal progenitor cells after a culture period; and (c) plotting the graft size against the expression level of the G gene to obtain data points for training samples. As illustrated herein, certain G genes are associated with the potential for transplantation and survival of neurons derived from neuronal progenitor cells to a certain size after implantation, as demonstrated by the use of rat models as implantation alternatives. Exemplary genes are identified through a machine learning process trained on reference neuronal progenitor cell populations based on gene expression, which are associated with the generation of grafts of a certain size after implantation into brain regions.

[0145] In some implementations, the reference plot is obtained by applying the gene expression levels of one or more G genes in the test sample as input to a machine learning model configured to predict whether neurons derived from neuronal progenitor cells will successfully transplant and survive after a population of neuronal progenitor cells is implanted into a brain region, wherein the machine learning model is trained using the gene expression levels of G genes in multiple reference populations of neuronal progenitor cells.

[0146] In some implementations, the transplant viability of the reference population indicates whether the reference population has successfully transplanted into the brain region of the subject after implantation. In some implementations, the transplant viability of the reference population indicates the extent to which the reference population has successfully transplanted into the brain region of the subject after implantation.

[0147] In some embodiments, the reference population includes a population of neuronal progenitor cells that survived transplantation after implantation. In some embodiments, the reference population includes a population of neuronal progenitor cells that did not survive transplantation after implantation. In some embodiments, the reference population includes both a population of neuronal progenitor cells that survived transplantation after implantation and a population of neuronal progenitor cells that did not survive transplantation after implantation.

[0148] In some embodiments, the reference population is derived from pluripotent stem cells under conditions that induce cellular neural differentiation. In some embodiments, the reference population has been cultured to differentiate the cells into specific dopaminergic neuronal progenitor cells. In some embodiments, the reference population comprises specific dopaminergic neuronal progenitor cells. Exemplary methods for inducing cellular neural differentiation to form the reference population are described herein. In some embodiments, the reference population is formed entirely using the same method for inducing cellular neural differentiation (e.g., any of the methods described herein). In some embodiments, the reference population is formed using multiple different methods for inducing cellular neural differentiation (e.g., any of the multiple methods described herein).

[0149] In some embodiments, the transplant viability adaptation of the reference population is determined based on the number of cells derived from a reference population present in the brain region after implantation. In some embodiments, cell counts are performed at the following times post-implantation: day 7, day 14, or day 21; approximately day 7, approximately day 14, or approximately day 21; at least day 7, at least day 14, or at least day 21; or at least approximately day 7, at least approximately day 14, or at least approximately day 21. In some embodiments, cell counts are performed on day 7, approximately day 7, at least day 7, or at least approximately day 7 post-implantation. In some embodiments, cell counts are performed on day 14, approximately day 14, at least day 14, or at least approximately day 14 post-implantation. In some embodiments, cell counts are performed on day 21, approximately day 21, at least day 21, or at least approximately day 21 post-implantation.

[0150] In some embodiments, a population of neuronal progenitor cells is predicted to produce transplanted, viable neurons in the brain region if cells derived from the implanted neuronal progenitor cells are predicted to form a graft size equal to or greater than a threshold graft size value in the subject's brain region after implantation. In some embodiments, the threshold graft size value is equal to or greater than 1,000 cells in a cross-section of the brain region. In some embodiments, the brain region is the substantia nigra, and the cross-section represents approximately one-sixth of the substantia nigra.

[0151] In some implementations, a method for predicting the cell transplant viability of neurons derived from a population of neuronal progenitor cells includes: (a) determining the gene expression levels of one or more genes (G genes) associated with transplant viability in a test sample comprising a population of neuronal progenitor cells, wherein the G genes are selected from the group consisting of: AC000120.3, KRT77, TTR, PRR16, MEGF10, PDE3A, GDPD2, CMTM8, APOA1, CMTM7, CDHR3, CORIN, VTN, CPNE8, EFEMP1, CD47, SPARC, JAM2, CDO1, PLXDC2, DYNLL2, ITGA3, RPS6KL1, CHRNB2, SULT4A1, PTPN 3. LZTS1, RUNX1T1, TMEM145, EPHA10, CARMIL3, MANEAL, TMEM176B, MPP3, DRAXIN, ADDRB1, KIF26A, CELF5, CNTN2, ASPHD1, SVOP, ANGPT2, SLC22A15, SRRM3, GRIN2D, DACH2, CHST1, GRIN1, LHX5, and NOS2; and (b) comparing the gene expression level of each of the one or more G genes in the test sample with a control level of G gene expression to predict whether neurons derived from neuronal progenitor cells are likely to transplant and survive in the brain region after a population of neuronal progenitor cells is implanted into the brain region of the subject.

[0152] In some implementations, the predicted transplant viability is determined for at least 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, or 20 G genes, and the overall transplant viability prediction for the test sample is based on a combined assessment of the predicted transplant viability obtained for each of the two or more G genes. In some implementations, the combined assessment involves determining a mean or median predicted transplant viability.

[0153] In some embodiments, the G gene is selected from the group consisting of: TTR, PRR16, CMTM8, APOA1, CD47, CD01, KIR26A, and CNTN2. In some embodiments, the G gene is three or more G genes selected from the group consisting of: TTR, PRR16, CMTM8, APOA1, CD47, CD01, KIR26A, and CNTN2. In some embodiments, the one or more G genes are selected from the group consisting of: TTR, PRR16, and CD47. In some embodiments, the G gene is TTR, PRR16, and CD47.

[0154] In some embodiments, the present invention provides a method for predicting whether a population of neuronal progenitor cells is likely to successfully transplant and survive when implanted into a brain region, wherein such methods involve: (a) determining the gene expression levels of one or more genes (G genes) associated with transplant survival in a test sample comprising a population of neuronal progenitor cells, wherein the one or more G genes are selected from the group consisting of: AC000120.3, KRT77, TTR, PRR16, MEGF10, PDE3A, GDPD2, CMTM8, APOA1, CMTM7, CDHR3, CORIN, VTN, CPNE8, EFEMP1, CD47, SPARC, JAM2, CDO1, PLXDC2, DYNLL2, ITGA3, RPS6KL1, CHRNB2, SULT4A1, PTPN3, LZTS1, RUNX1T1, TMEM145 EPHA10, CARMIL3, MANEAL, TMEM176B, MPP3, DRAXIN, ADDRB1, KIF26A, CELF5, CNTN2, ASPHD1, SVOP, ANGPT2, SLC22A15, SRRM3, GRIN2D, DACH2, CHST1, GRIN1, LHX5, and NOS2; and (b) comparing the expression level of each of the one or more G genes in the tested neuronal progenitor population to a predetermined threshold for a specific G gene, wherein a high transplant viability of the neuronal progenitor is indicated if the expression level is either (i) above the predetermined threshold for the G gene; or (ii) below the predetermined threshold for the G gene; wherein “above” or “below” is defined by the known biological relevance of the G gene in terms of transplant viability.

[0155] In some of these embodiments, the predetermined threshold for a specific G gene is based on the expression level of the G gene in a training sample comprising neuronal progenitor cells known to exhibit high graft survival rates when implanted into the brain, and the gene expression level of the G gene in a test sample similar to that in the training sample predicts that neurons derived from the neuronal progenitor cells in the test sample have high graft survival potential. In other embodiments, the predetermined threshold for a specific G gene is based on the expression level of the G gene in a training sample comprising neuronal progenitor cells known to exhibit low graft survival rates when implanted into the brain, and the gene expression level of the G gene in a test sample similar to that in a control sample predicts that neurons derived from the neuronal progenitor cells in the test sample have low graft survival potential.

[0156] In some embodiments, prediction is made by comparing the gene expression level of each of the one or more G genes in the test sample with a control level of G gene expression. In some embodiments, the control expression level is obtained by performing a determination of gene expression level on the control sample. In some embodiments, the control sample comprises a known amount of nucleotides encoding at least a portion of the G gene. In some embodiments, the control sample comprises a reference sample of neuronal progenitor cells. In some embodiments, the reference sample comprises a pooled sample of neuronal progenitor cells from multiple donors. In some embodiments, the reference sample comprises neuronal progenitor cells known to produce neurons exhibiting high transplant viability levels, and the gene expression level of the G gene in the test sample, similar to that in the control sample, predicts high transplant viability potential for neurons derived from the neuronal progenitor cells. In some embodiments, the reference sample comprises neuronal progenitor cells known to produce neurons exhibiting low transplant viability levels, and the gene expression level of the G gene in the test sample, similar to that in the control sample, predicts low transplant viability potential for neurons derived from the neuronal progenitor cells.

[0157] In some embodiments, the control expression level is a predetermined threshold, such as a threshold expression level based on a positive or negative correlation with whether neurons derived from neuronal progenitors have high transplant viability in brain regions. In some embodiments, transplant viability of neurons derived from neuronal progenitors in brain regions is predicted if: (i) the gene expression level of at least one first G gene selected from the group consisting of: AC000120.3, KRT77, TTR, PRR16, MEGF10, PDE3A, GDPD2, CMTM8, APOA1, CMTM7, CDHR3, CORIN, VTN, CPNE8, EFEMP1, CD47, SPARC, JAM2, CDO1, and PLXDC2; and / or (ii) the group consisting of: The gene expression level of at least one second G gene is greater than a predetermined threshold for the second G gene: DYNLL2, ITGA3, RPS6KL1, CHRNB2, SULT4A1, PTPN3, LZTS1, RUNX1T1, TMEM145, EPHA10, CARMIL3, MANEAL, TMEM176B, MPP3, DRAXIN, ADGRB1, KIF26A, CELF5, CNTN2, ASPHD1, SVOP, ANGPT2, SLC22A15, SRRM3, GRIN2D, DACH2, CHST1, GRIN1, LHX5, and NOS2.

[0158] In some embodiments, the gene expression level is the ratio of the relative expression levels of the G gene and the reference gene in the test sample, and a predetermined threshold is a threshold for this ratio. In some embodiments, each ratio is calculated as counts per million (CPM) of [gene] / CPM of the reference gene. In some embodiments, each ratio is calculated as log CPM of [gene] / log CPM of the reference gene. In some embodiments, each ratio is calculated by qPCR as the relative expression of the target gene relative to a reference gene (such as a housekeeping gene). In some embodiments, the reference gene is a housekeeping gene. In some implementations, the reference gene is selected from PRS18, IPO8, RPL113A, HSP90AB1, UBC, PSMC4, SDHA, HPRT1, HMBS, TFRC, PPIA, RPL30, GUSB, ACTB, LDHA, RPS17, GAPDH, PPIH, NONO, PUM1, HBB, G6PD, TBP, ALAS1, PGK1, CDKN1A, YWHAZ, POP4, RPLP0, or B2M. In some implementations, the reference gene is a housekeeping gene, such as GAPDH.

[0159] In some implementations, a predetermined threshold for a specific G gene is based on the ratio of the relative expression levels of a) the G gene and b) the control gene in the test sample. For example, in some implementations, the control gene is GAPDH, and the pre-determined thresholds are selected from a group consisting of: (a) a ratio of AC000120.3 to GAPDH expression less than about 0.14; (b) a ratio of KRT77 to GAPDH expression less than about 0.68; (c) a ratio of TTR to GAPDH expression less than about 1.11; (d) a ratio of PRR16 to GAPDH expression less than about 0.43; (e) a ratio of MEGF10 to GAPDH expression less than about 0.79; (f) a ratio of PDE3A to GAPDH expression less than about 1.00; (g) a ratio of GDPD2 to GAPDH expression less than about 0.78; (h) a ratio of CMTM8 to GAPDH expression less than about 1.02; (i) a ratio of APOA1 to GAPDH expression less than about 0.68; (j) (k) The ratio of CMTM7 to GAPDH expression less than about 0.88; (l) The ratio of CDHR3 to GAPDH expression less than about 1.09; (m) The ratio of CORIN to GAPDH expression less than about 1.24; (n) The ratio of VTN to GAPDH expression less than about 0.98; (o) The ratio of CPNE8 to GAPDH expression less than about 0.79; (p) The ratio of EFEMP1 to GAPDH expression less than about 0.83; (q) The ratio of CD47 to GAPDH expression less than about 1.16; (r) The ratio of SPARC to GAPDH expression less than about 1.29; (s) The ratio of JAM2 to GAPDH expression less than about 0.82; (t) The ratio of CDO1 to GAPDH expression less than about 1.00; (v) The ratio of PLXDC2 to GAPDH expression less than about 1.00; (v) (v) Ratio of DYNLL2 to GAPDH expression greater than approximately 0.56; (w) Ratio of ITGA3 to GAPDH expression greater than approximately 0.26; (x) Ratio of RPS6KL1 to GAPDH expression greater than approximately 0.21; (y) Ratio of CHRNB2 to GAPDH expression greater than approximately 0.23; (z) Ratio of SULT4A1 to GAPDH expression greater than approximately 0.22; (aa) Ratio of PTPN3 to GAPDH expression greater than approximately 0.03; (aa) Ratio of LZTS1 to GAPDH expression greater than approximately 0.19; (ab) Ratio of RUNX1T1 to GAPDH expression greater than approximately 0.24; (ac) Ratio of TMEM145 to GAPDH expression greater than approximately 0.05; (ad) Ratio of EPHA10 to GAPDH expression greater than approximately 0.The ratios of CARMIL3 and GAPDH expression were as follows: (ae) CARMIL3 to GAPDH expression greater than approximately 0.16; (af) MANEAL to GAPDH expression greater than approximately 0.24; (ag) TMEM176B to GAPDH expression greater than approximately 0.11; (ah) MPP3 to GAPDH expression greater than approximately 0.12; (ai) DRAXIN to GAPDH expression greater than approximately 0.27; (aj) ADGRB1 to GAPDH expression greater than approximately 0.07; (ak) KIF26A to GAPDH expression greater than approximately 0.23; (al) CELF5 to GAPDH expression greater than approximately 0.25; (am) CNTN2 to GAPDH expression greater than approximately 0.23; (an) ASPHD1 to GAPDH expression greater than approximately 0.08; (ao) The ratios of SVOP to GAPDH expression greater than approximately 0.16; (ap) ANGPT2 to GAPDH expression greater than approximately 0.06; (aq) SLC22A15 to GAPDH expression greater than approximately 0.04; (ar) SRRM3 to GAPDH expression greater than approximately 0.17; (as) GRIN2D to GAPDH expression greater than approximately 0.02; (at) DACH2 to GAPDH expression greater than approximately 0.06; (au) CHST1 to GAPDH expression greater than approximately 0.04; (av) GRIN1 to GAPDH expression greater than approximately 0.26; (aw) LHX5 to GAPDH expression greater than approximately 0.06; and (ax) NOS2 to GAPDH expression greater than approximately 0.08.

[0160] In some embodiments, the predetermined thresholds are selected from the group consisting of: (i) a ratio of TTR to GAPDH expression less than about 1.11; (ii) a ratio of PRR16 to GAPDH expression less than about 0.43; (iii) a ratio of CMTM8 to GAPDH expression less than about 1.02; (iv) a ratio of APOA1 to GAPDH expression less than about 0.68; (v) a ratio of CD47 to GAPDH expression less than about 1.16; (vi) a ratio of CDO1 to GAPDH expression less than about 1.00; (vii) a ratio of KIF26A to GAPDH expression greater than about 0.23; and (viii) a ratio of CNTN2 to GAPDH expression greater than about 0.23. In some embodiments, the predetermined thresholds are selected from at least two, three, four, five, six, or seven ratios from (i) to (viii).

[0161] In some implementations, the predetermined first threshold level is selected from: a ratio of TTR to GAPDH expression of less than about 1.11; and / or a ratio of PRR16 to GAPDH expression of less than about 0.43; and / or a ratio of CD47 to GAPDH expression of less than about 1.16.

[0162] In some implementations, the predetermined first thresholds are: a ratio of TTR to GAPDH expression less than about 1.11; a ratio of PRR16 to GAPDH expression less than about 0.43; and a ratio of CD47 to GAPDH expression less than about 1.16.

[0163] In some embodiments, the present invention also provides methods for training a machine learning model to predict whether a population of neuronal progenitor cells is likely to successfully transplant and survive when implanted into a brain region. These methods may include: (a) obtaining gene expression levels of one or more genes in each of a plurality of reference populations of neuronal progenitor cells; (b) receiving transplantation survival adaptation information for each of the plurality of reference populations, wherein the transplantation survival adaptation information of the reference populations indicates whether or to what extent the neuronal progenitor cells have transplanted and survived in the brain region after implantation of the reference populations of neuronal progenitor cells into a brain region of a subject; and (c) applying the gene expression levels of (a) and the transplantation survival adaptation information of (b) as inputs to train a machine learning model, wherein the machine learning model is trained to predict, based on the gene expression levels of multiple genes, whether the neuronal progenitor cell population will transplant and survive in the brain region after implantation into a brain region of a subject.

[0164] Predicting dopamine production In some embodiments, the present invention provides a method for predicting whether neurons derived from a population of neuronal progenitor cells will produce dopamine. In some embodiments, these methods involve predicting whether neurons generated or derived from neuronal progenitor cells will produce dopamine, such as a threshold level of dopamine. In some embodiments, the prediction is based on the cell's ability to produce a specific amount of dopamine, such as determined on a per-cell basis.

[0165] In some implementations, the gene expression level of one of a plurality of genes in a test sample of a neuronal progenitor cell population is used to predict dopamine production in cells derived from neuronal progenitor cells, wherein the gene is associated with the ability of neurons derived from neuronal progenitor cells to produce or release dopamine (hereinafter referred to as the “D gene”). Exemplary D genes are described herein and are included in Table E2. In some implementations, the provided method involves: (a) determining the gene expression levels of one or more genes (D genes) associated with predicted dopamine production in a test sample comprising a population of neuronal progenitor cells, wherein the D genes are selected from the group consisting of: CNTNAP5, KLHL1, NHLH2, GREM2, BRINP2, GRIN3A, LRRC4C, IRX3, CPNE4, PTPN3, PMEL, PCDH20, LRRC37A2, TMEM246, B3GALNT1, ZHX1, BCAS4, SLC25A37, GRINA, MID1, FRMD4A, PARP10, WHAMMP2, EYA1, CORO2B, WHAMMP3, B3GALT5, GPR35, ABCD2, IT IH3, AC107464.1, CAMK2N1, CAMK2A, PRPS1, GOLGA6L10, AMOT, SULT1A1, CD83, SPON1, FRMPD3, AC096570.1, TCAF2, GOLGA8M, VWA5B2, CA8, AC017050.1, KRT77, AP000350.6, LINC02751, and ARHGAP5-AS1; and (b) predicting the dopamine-producing capacity of neurons derived from neuronal progenitor cells by associating the determined gene expression levels of the one or more D genes in the test samples with a reference map of each D gene, the reference map associating neuronal dopamine production with the gene expression levels of the D genes in a training set including one or more reference samples. In some embodiments, a population of neuronal progenitor cells for implantation for therapeutic purposes is selected based on the gene expression levels of one or more D genes.

[0166] In some embodiments, each data point on the reference plot is determined by the following steps: (a) measuring the gene expression level of the D gene in a reference sample comprising a population of neuronal progenitor cells; (b) differentiating the neuronal progenitor cells to generate neurons and measuring the amount of dopamine produced by the neurons derived from the neuronal progenitor cells; and (c) plotting dopamine production against the expression level of the D gene to obtain data points for training samples. In some embodiments, the reference plot includes multiple data points obtained for each of a plurality of reference samples.

[0167] In some implementations, the reference plot is obtained by applying the gene expression levels of one or more D genes in the test sample as input to a machine learning model configured to predict whether neurons derived from a population of neuronal progenitor cells will produce dopamine, wherein the machine learning model is trained using the gene expression levels of the D gene in multiple reference populations of neuronal progenitor cells.

[0168] As illustrated herein, certain D genes are associated with the potential of neuronal progenitor cells to generate dopamine-producing neurons, as demonstrated by alternative assays involving long-term in vitro culture of neuronal progenitor cells under conditions in which cells mature into dopaminergic neurons in culture. Specifically, long-term culture is a formulation for preparing mature dopamine neurons in vitro by culturing neuronal progenitor cells for approximately 60 days (including initial culture to differentiate iPSCs into neuronal progenitor cells, and an additional 60 days of culture to generate mature dopamine neurons). Exemplary genes are identified through a machine learning process trained on gene expression based on a population of reference neuronal progenitor cells that are associated with specific threshold amounts of dopamine production after culture to prepare mature dopamine neurons. In some embodiments, in addition to dopamine neurotransmitter data, a metric of "serotonin" may be determined as an alternative to or supplement to a determined metric of dopamine production in the culture formulation of mature dopamine neurons. In some embodiments, dopamine production and serotonin production may be assessed after stimulation of mature dopamine neurons with potassium chloride (KCl). In some implementations, neurotransmitter release can be assessed by liquid chromatography-mass spectrometry (LC-MS).

[0169] In some embodiments, dopamine release from a reference population of mature dopamine neurons derived from a reference neuronal progenitor cell population indicates whether the reference population is capable of generating dopamine-producing neurons. In some embodiments, dopamine production adaptation of the reference population indicates the extent to which the reference population is capable of generating dopamine-producing neurons.

[0170] In some embodiments, the reference population includes a population of neuronal progenitor cells that differentiate into mature dopaminergic neurons and produce dopamine. In some embodiments, the reference population includes a population of neuronal progenitor cells that differentiate into neurons that do not produce dopamine. In some embodiments, the reference population includes both a population of neuronal progenitor cells that differentiate into mature dopaminergic neurons and produce dopamine, and a population of neuronal progenitor cells that differentiate into neurons that do not produce dopamine.

[0171] In some embodiments, if it is predicted that neurons derived from implanted neuronal progenitor cells will produce dopamine at or above a dopamine threshold, then the neuronal progenitor cell population is predicted to produce cells exhibiting dopamine production. In some embodiments, the dopamine threshold is equal to or greater than 15 nM dopamine / 10 5 Each cell.

[0172] In some implementations, methods for predicting dopamine production in neurons derived from a population of neuronal progenitor cells include: (a) determining the gene expression levels of one or more genes (D genes) associated with dopamine production in a test sample comprising a population of neuronal progenitor cells, wherein the D genes are selected from the group consisting of: CNTNAP5, KLHL1, NHLH2, GREM2, BRINP2, GRIN3A, LRRC4C, IRX3, CPNE4, PTPN3, PMEL, PCDH20, LRRC37A2, TMEM246, B3GALNT1, ZHX1, BCAS4, SLC25A37, GRINA, MID1, FRMD4A, PARP10, WHAMMP2, EYA1, CORO2B, WHAMMP3, B3GALT5, GPR35, ABCD2, ITIH3, AC107464.1 (a) CAMK2N1, CAMK2A, PRPS1, GOLGA6L10, AMOT, SULT1A1, CD83, SPON1, FRMPD3, AC096570.1, TCAF2, GOLGA8M, VWA5B2, CA8, AC017050.1, KRT77, AP000350.6, LINC02751, and ARHGAP5-AS1; and (b) comparing the gene expression level of each of the one or more D genes in the test sample with a predetermined threshold for a specific D gene, wherein a high transplant viability of neuronal progenitor cells is indicated if the expression level is either (i) above the predetermined threshold for the D gene; or (ii) below the predetermined threshold for the D gene; wherein “above” or “below” is defined by the known biological relevance of the D gene in terms of dopamine production capacity.

[0173] In some embodiments of these implementations, a predetermined threshold for a specific D gene is based on the expression level of the D gene in training samples, which include neuronal progenitor cells known to generate neurons that produce large amounts of dopamine, and the gene expression level of the D gene in test samples, which is similar to the expression level of the D gene in the training samples, predicts that neurons derived from the neuronal progenitor cells in the test samples have dopamine-producing potential. In other embodiments, the predetermined threshold for a specific D gene is based on the expression level of the D gene in training samples, which include neuronal progenitor cells known to generate neurons that produce small amounts of dopamine, and the gene expression level of the D gene in test samples, which is similar to the expression level of the D gene in control samples, predicts that neurons derived from the neuronal progenitor cells in the test samples have low dopamine-producing potential.

[0174] In some embodiments, dopamine production capacity is determined against at least 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, or 20 D genes, and the overall dopamine production capacity prediction for a test sample is based on a combined assessment of the predicted dopamine production capacity obtained for each of the two or more D genes. In some embodiments, the combined assessment involves determining a mean or median predicted dopamine production capacity.

[0175] In some implementations, the D gene is selected from the group consisting of: CNTNAP5, NHLH2, GREM2, PMEL, PCDH20, LRRC37A2, SLC25A37, MID1, EYA1, B3GALT5, GPR35, AC107464.1, CAMK2N1, CAMK2A, GOLGA6L10, FRMPD3, VWA5B2, AC017050.1, and LINC02751. In some embodiments, the D gene is three or more D genes selected from the group consisting of: CNTNAP5, NHLH2, GREM2, PMEL, PCDH20, LRRC37A2, SLC25A37, MID1, EYA1, B3GALT5, GPR35, AC107464.1, CAMK2N1, CAMK2A, GOLGA6L10, FRMPD3, VWA5B2, AC017050.1, and LINC02751. In some embodiments, the one or more D genes are B3GALT5, FRMPD3, and GREM2.

[0176] In some embodiments, prediction is made by comparing the gene expression level of each of the one or more D genes in the test sample with a control level of D gene expression. In some embodiments, the control expression level is obtained by performing a determination of gene expression level on the control sample. In some embodiments, the control sample comprises a known amount of nucleotides encoding at least a portion of the D gene. In some embodiments, the control sample comprises a reference sample of neuronal progenitor cells. In some embodiments, the reference sample comprises a pooled sample of neuronal progenitor cells from multiple donors. In some embodiments, the reference sample comprises neuronal progenitor cells known to produce neurons exhibiting high dopamine production levels, and the gene expression level of the D gene in the test sample, similar to that in the control sample, predicts that neurons derived from the neuronal progenitor cells have high dopamine production potential. In some embodiments, the reference sample comprises neuronal progenitor cells known to produce neurons exhibiting low dopamine production levels, and the gene expression level of the D gene in the test sample, similar to that in the control sample, predicts that neurons derived from the neuronal progenitor cells have low dopamine production potential.

[0177] In some embodiments, the control expression level is a predetermined threshold, such as a threshold expression level based on whether it is positively or negatively correlated with the likelihood of dopamine production by neurons derived from neuronal progenitors. In some embodiments, dopamine production by neurons derived from neuronal progenitors is predicted if: (i) the gene expression level of at least one first D gene selected from the group consisting of: CNTNAP5, KLHL1, NHLH2, GREM2, BRINP2, GRIN3A, LRRC4C, IRX3, CPNE4, PTPN3, PMEL, PCDH20, LRRC37A2, TMEM246, B3GALNT1, and ZHX1; and / or (ii) the gene expression level of at least one second D gene selected from the group consisting of: BCAS 4. SLC25A37, GRINA, MID1, FRMD4A, PARP10, WHAMMP2, EYA1, CORO2B, WHAMMP3, B3GALT5, GPR35, ABCD2, ITIH3, AC107464.1, CAMK2N1, CAMK2A, PRPS1, GO LGA6L10, AMOT, SULT1A1, CD83, SPON1, FRMPD3, AC096570.1, TCAF2, GOLGA8M, VWA5B2, CA8, AC017050.1, KRT77, AP000350.6, LINC02751 and ARHGAP5-AS1.

[0178] In some embodiments, the gene expression level is the ratio of the relative expression levels of gene D in the test sample to that of a reference gene, and a predetermined threshold is a threshold for this ratio. In some embodiments, each ratio is calculated as counts per million (CPM) of [gene] / CPM of the reference gene. In some embodiments, each ratio is calculated as log CPM of [gene] / log CPM of the reference gene. In some embodiments, each ratio is calculated by qPCR as the relative expression of the target gene relative to a reference gene (such as a housekeeping gene). In some embodiments, the reference gene is a housekeeping gene. In some implementations, the reference gene is selected from PRS18, IPO8, RPL113A, HSP90AB1, UBC, PSMC4, SDHA, HPRT1, HMBS, TFRC, PPIA, RPL30, GUSB, ACTB, LDHA, RPS17, GAPDH, PPIH, NONO, PUM1, HBB, G6PD, TBP, ALAS1, PGK1, CDKN1A, YWHAZ, POP4, RPLP0, or B2M. In some implementations, the reference gene is a housekeeping gene, such as GAPDH.

[0179] In some implementations, a predetermined threshold for a specific D gene is based on the ratio of the relative expression levels of a) the D gene and b) the control gene in the test sample. For example, in some implementations, the control gene is GAPDH, and the pre-determined threshold is selected from a group consisting of: (a) a ratio of CNTNAP5 to GAPDH expression less than about 0.12; (b) a ratio of KLHL1 to GAPDH expression less than about 0.10; (c) a ratio of NHLH2 to GAPDH expression less than about 0.56; (d) a ratio of GREM2 to GAPDH expression less than about 0.35; (e) a ratio of BRINP2 to GAPDH expression less than about 0.97; (f) a ratio of GRIN3A to GAPDH expression less than about 0.48; (g) a ratio of LRRC4C to GAPDH expression less than about 0.39; (h) a ratio of IRX3 to GAPDH expression less than about 0.55; (i) a ratio of CPNE4 to GAPDH expression less than about 0.28; (j) a ratio of PTPN3 to GAPDH expression less than about 0.25; (k) The following ratios were observed: (l) PMEL to GAPDH expression ratio less than approximately 0.29; (m) PCDH20 to GAPDH expression ratio less than approximately 0.20; (n) LRRC37A2 to GAPDH expression ratio less than approximately 0.68; (n) TMEM246 to GAPDH expression ratio less than approximately 0.53; (o) B3GALNT1 to GAPDH expression ratio less than approximately 0.67; (p) ZHX1 to GAPDH expression ratio less than approximately 0.55; (q) BCAS4 to GAPDH expression ratio greater than approximately 0.42; (r) SLC25A37 to GAPDH expression ratio greater than approximately 0.38; (s) GRINA to GAPDH expression ratio greater than approximately 0.60; (t) MID1 to GAPDH expression ratio greater than approximately 0.62; (u) The ratio of FRMD4A to GAPDH expression greater than approximately 0.57; (v) the ratio of PARP10 to GAPDH expression greater than approximately 0.25; (w) the ratio of WHAMMP2 to GAPDH expression greater than approximately 0.37; (x) the ratio of EYA1 to GAPDH expression greater than approximately 0.32; (y) the ratio of CORO2B to GAPDH expression greater than approximately 0.40; (z) the ratio of WHAMMP3 to GAPDH expression greater than approximately 0.34; (aa) the ratio of B3GALT5 to GAPDH expression greater than approximately 0.40; (ab) the ratio of GPR35 to GAPDH expression greater than approximately 0.19; (ac) the ratio of ABCD2 to GAPDH expression greater than approximately 0.35; (ad) the ratio of ITIH3 to GAPDH expression greater than approximately 0.The ratios of 17; (ae) AC107464.1 to GAPDH expression greater than approximately 0.20; (af) CAMK2N1 to GAPDH expression greater than approximately 0.52; (ag) CAMK2A to GAPDH expression greater than approximately 0.37; (ah) PRPS1 to GAPDH expression greater than approximately 0.52; (ai) GOLGA6L10 to GAPDH expression greater than approximately 0.21; (aj) AMOT to GAPDH expression greater than approximately 0.50; (ak) SULT1A1 to GAPDH expression greater than approximately 0.18; (al) CD83 to GAPDH expression greater than approximately 0.29; (am) SPON1 to GAPDH expression greater than approximately 0.76; (an) FRMPD3 to GAPDH expression greater than approximately 0.31; (ao) The ratio of AC096570.1 to GAPDH expression greater than approximately 0.14; (ap) TCAF2 to GAPDH expression greater than approximately 0.30; (aq) GOLGA8M to GAPDH expression greater than approximately 0.003; (ar) VWA5B2 to GAPDH expression greater than approximately 0.22; (as) CA8 to GAPDH expression greater than approximately 0.19; (at) AC017050.1 to GAPDH expression greater than approximately 0.08; (au) KRT77 to GAPDH expression greater than approximately 0.14; (av) AP000350.6 to GAPDH expression greater than approximately 0.31; (aw) LINC02751 to GAPDH expression greater than approximately 0.19; and (ax) ARHGAP5-AS1 to GAPDH expression greater than approximately 0.26.

[0180] In some implementations, the predetermined thresholds are selected from the group consisting of: (i) a ratio of CNTNAP5 to GAPDH expression less than about 0.12; (ii) a ratio of NHLH2 to GAPDH expression less than about 0.56; (iii) a ratio of GREM2 to GAPDH expression less than about 0.35; (iv) a ratio of PMEL to GAPDH expression less than about 0.29; (v) a ratio of PCDH20 to GAPDH expression less than about 0.20; (vi) a ratio of LRRC37A2 to GAPDH expression less than about 0.68; (vii) a ratio of SLC25A37 to GAPDH expression greater than about 0.38; (viii) a ratio of MID1 to GAPDH expression greater than about 0.62; (ix) a ratio of EYA1 to GAPDH expression greater than about 0.32; (x) The following ratios were observed: (xi) B3GALT5 to GAPDH expression greater than approximately 0.40; (xii) GPR35 to GAPDH expression greater than approximately 0.19; (xii) AC107464.1 to GAPDH expression greater than approximately 0.20; (xiii) CAMK2N1 to GAPDH expression greater than approximately 0.52; (xiv) CAMK2A to GAPDH expression greater than approximately 0.37; (xv) GOLGA6L10 to GAPDH expression greater than approximately 0.21; (xvi) FRMPD3 to GAPDH expression greater than approximately 0.31; (xvii) VWA5B2 to GAPDH expression greater than approximately 0.22; (xviii) AC017050.1 to GAPDH expression greater than approximately 0.08; and (xix). The ratio of LINC02751 to GAPDH expression is greater than about 0.19. In some embodiments, a predetermined threshold is selected from at least two, three, four, five, six, seven, eight, nine, or ten ratios from (i) to (xix).

[0181] In some implementations, the predetermined first threshold level is selected from: a ratio of SLC25A37 to GAPDH expression greater than about 0.38; and / or a ratio of GPR35 to GAPDH expression greater than about 0.19; and / or a ratio of CAMK2N1 to GAPDH expression greater than about 0.52.

[0182] In some implementations, the predetermined first thresholds are: a ratio of SLC25A37 to GAPDH expression greater than about 0.38; a ratio of GPR35 to GAPDH expression greater than about 0.19; and a ratio of CAMK2N1 to GAPDH expression greater than about 0.52.

[0183] In some embodiments, the present invention provides methods for training machine learning models to predict whether neurons derived from a population of neuronal progenitor cells will produce dopamine. These methods may include: (a) obtaining gene expression levels of one or more genes in each of a plurality of reference populations of neuronal progenitor cells; (b) receiving dopamine production information of neurons derived from each of the plurality of reference populations, wherein the dopamine production information of the reference populations indicates whether or to what extent cells derived from neuronal progenitor cells produce dopamine; and (c) applying the gene expression levels of (a) and the dopamine production information of (b) as input to train a machine learning model, wherein the machine learning model is trained to predict whether neurons derived from a population of neuronal progenitor cells will produce dopamine based on the gene expression levels of a plurality of genes.

[0184] Efficacy Measurement Matrix In some embodiments, the method provided by this invention can help select specific batches of in vitro neuronal progenitor cells that may be effective in treating neurodegenerative diseases such as Parkinson's disease. In some embodiments, the provided methods for predicting successful transplant survival and dopamine production can be used as part of a power assay matrix. In some embodiments, the power assay matrix may also include additional assays.

[0185] In some implementations, in addition to testing for predicted transplant viability and / or dopamine production, these methods also include assessing whether the cell population has the desired differentiation state, such as the differentiation state of neuronal progenitors (e.g., identified dopaminergic neuronal progenitors). In some embodiments, these methods include any of the methods described in the following patent documents: PCT / US2020 / 043627 entitled “METHODS OF IDENTIFYING DOPAMINERGIC NEURONS AND PROGENITOR CELLS”, PCT / US2022 / 073974 entitled “METHODS OF DIFFERENTIATING NEURALCELLS AND PREDICTING ENGRAFTMENT THEREOF”, and U.S. Patent Publication 2023 / 0377685 entitled “METHODS OF CLASSIFING THE DIFFERENTIATION STATE OF CELLS AND RELATED COMPOSITIONS OF DIFFERENTIATED CELLS”, the entire contents of which are incorporated herein by reference.

[0186] In some embodiments, the present invention provides a potency assay matrix for determining the efficacy of neuronal progenitor cell populations in treating neurodegenerative diseases. The potency assay matrix includes a method of subjecting the neuronal progenitor cell population to at least two of the steps (a), (b), and (c): (a) Expected differentiation state : Classify an in vitro population of neuronal progenitor cells to determine whether neuronal progenitor cells include identifiable dopaminergic precursor cells by the following steps: (i) receiving a test dataset as input, the test dataset including the expression levels of one or more genes expressed in a first test sample including neuronal progenitor cells; (ii) using the test dataset and a first reference dataset to calculate a first similarity score for the first test sample, wherein: (1) the first reference dataset includes a representation of the gene expression levels of one or more genes differentially expressed between cells in a first differentiation state and cells in a second differentiation state, wherein the second differentiation state is the differentiation state of identifiable dopaminergic neurons, and wherein (i) The first differentiation state is earlier or later than the second differentiation state in the stem cell differentiation pathway; (ii) The expression levels in the test dataset include the expression levels of one or more genes included in the first reference dataset; and (iii) The first similarity score indicates whether the differentiation state of the test cells is more similar to the first or second differentiation state; (iv) The novelty score of the neuronal progenitor cells in the first test sample is determined, wherein the novelty score indicates the degree of deviation of the gene expression levels in the test dataset from the gene expression levels in the reference database; and (v) Based on the similarity score and the novelty score, it is determined whether the first test sample includes the identified dopaminergic neuronal cells. (b) Transplant survivalThe following steps are used to predict the likelihood of successful transplantation survival of neuronal progenitor cells when implanted into a brain region: (i) Determine the gene expression levels of one or more genes (G genes) associated with the predicted transplantation survival potential in a second test sample including neuronal progenitor cells, wherein the one or more G genes are selected from the group consisting of: AC000120.3, KRT77, TTR, PRR16, MEGF10, PDE3A, GDPD2, CMTM8, APOA1, CMTM7, CDHR3, CORIN, VTN, CPNE8, EFEMP1, CD47, SPARC, JAM2, CDO1, PLXDC2, DYNLL2, ITGA3, RPS6KL1, CHRNB2, SULT4A1, PTPN3, LZTS1, RUNX1T1, TMEM145, EPHA10, CARMIL3, MANEAL, TMEM176B, MPP3, DRAXIN, ADDRB1, KIF26A, CELF5, CNTN2, ASPHD1, SVOP, ANGPT2, SLC22A15, SRRM3, GRIN2D, DACH2, CHST1, GRIN1, LHX5, and NOS2; and (ii) predicting neuronal transplant viability of neuronal progenitor cells by correlating the determined gene expression levels of one or more G genes in a second test sample with a reference map of each G gene, the reference map correlating graft size with gene expression levels of G genes in a training set including one or more reference samples; and (c) Dopamine productionThe following steps are used to predict whether neurons derived from a population of neuronal progenitor cells will produce dopamine: (i) Determine the gene expression levels of one or more genes (D genes) associated with predicted dopamine production in a third test sample comprising the neuronal progenitor cell population, wherein the D genes are selected from the group consisting of: CNTNAP5, KLHL1, NHLH2, GREM2, BRINP2, GRIN3A, LRRC4C, IRX3, CPNE4, PTPN3, PMEL, PCDH20, LRRC37A2, TMEM246, B3GALNT1, ZHX1, BCAS4, SLC25A37, GRINA, MID1, FRMD4A, PARP10, WHAMMP2, EYA1, CORO2B, WHAMMP3, B3GALT5, GPR35, ABC D2, ITIH3, AC107464.1, CAMK2N1, CAMK2A, PRPS1, GOLGA6L10, AMOT, SULT1A1, CD83, SPON1, FRMPD3, AC096570.1, TCAF2, GOLGA8M, VWA5B2, CA8, AC017050.1, KRT77, AP000350.6, LINC02751, and ARHGAP5-AS1; and (ii) predicting the dopamine-producing capacity of neurons derived from neuronal progenitors by associating the determined gene expression levels of the one or more D genes in a third test sample with a reference map of each D gene, the reference map associating neuronal dopamine production with the gene expression levels of the D genes in a training set including one or more reference samples of neuronal progenitors.

[0187] In some embodiments, the power matrix determination includes at least two of the following: (a) predicting successful transplant survival; (b) predicting dopamine production; and (c) determining the differentiation status of a neuronal progenitor cell population to determine the suitability of the neuronal progenitor cell population for therapeutic use. In some embodiments, the power matrix determination includes methods (a) and (b). In some embodiments, the power matrix determination includes methods (a) and (c). In some embodiments, the power matrix determination includes methods (b) and (c). And in some embodiments, the power matrix determination includes all three steps of (a), (b), and (c).

[0188] In some embodiments, the power assay matrix includes step (b), and the G gene is selected from the group consisting of: TTR, PRR16, CMTM8, APOA1, CD47, CD01, KIR26A, and CNTN2. In some embodiments, the one or more G genes are TTR, PRR16, and CD47.

[0189] In some embodiments, the power assay matrix includes step (c), and the one or more D genes are selected from the group consisting of: CNTNAP5, NHLH2, GREM2, PMEL, PCDH20, LRRC37A2, SLC25A37, MID1, EYA1, B3GALT5, GPR35, AC107464.1, CAMK2N1, CAMK2A, GOLGA6L10, FRMPD3, VWA5B2, AC017050.1, and LINC02751. In some embodiments, the one or more D genes are B3GALT5, FRMPD3, and GREM2.

[0190] In some embodiments, these methods involve determining the expression levels of both G and D genes to identify neuronal progenitor cells predicted to produce cells capable of both transplant survival and dopamine production. In some embodiments, the gene expression levels of the G and D genes are associated, respectively, with the ability of the neuronal progenitor cell population to successfully form a graft after implantation into a brain region of a subject, and with the ability of the neuronal progenitor cell population to generate dopamine-producing neurons. In some embodiments, the neuronal progenitor cell population for implantation is selected based on the gene expression levels of one or more G genes and one or more D genes.

[0191] In some embodiments, these methods can be performed using any method capable of assessing gene expression levels. In some embodiments, these methods involve PCR analysis (e.g., qPCR) of one or more genes (e.g., G genes and / or D genes) in a neuronal progenitor cell population. In some embodiments, these methods involve RNA-seq analysis of one or more genes (e.g., G genes and / or D genes) in a neuronal progenitor cell population. In some embodiments, one or more G genes and one or more D genes are assessed using these methods. In some embodiments, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, or more genes are assessed. In some embodiments, the provided methods involve PCR-based methods for assessing 2, 3, 4, 5, or 6 genes. In some embodiments, the PCR may be multiplex PCR. In some embodiments, the PCR may also include gene expression levels of housekeeping genes (e.g., GAPDH) to facilitate normalization of gene expression levels. In some embodiments, gene expression levels of test samples can be measured, such as through RNA-seq analysis, and expression levels can be provided in the dataset as input to a process configured to make predictions using a machine learning model trained on the gene expression levels of the corresponding genes. In some embodiments, the process is configured to predict whether neurons derived from neuronal progenitor cells are likely to transplant successfully in a brain region after implantation of a population of neuronal progenitor cells into a subject, wherein the machine learning model is trained using the gene expression levels of the G gene from multiple reference populations of neuronal progenitor cells. In some embodiments, the process is configured to predict whether neurons derived from neuronal progenitor cells are capable of dopamine production, such as after implantation of a population of neuronal progenitor cells into a brain region, wherein the machine learning model is trained using the gene expression levels of the D gene from multiple reference populations of neuronal progenitor cells. In some embodiments, the process can be configured to predict both transplant viability and dopamine production capacity based on the G gene and the D gene, respectively.

[0192] Machine learning models and computing devices Exemplary machine learning model In some embodiments, the provided method includes predicting whether a population of neuronal progenitor cells possesses a characteristic associated with function, activity, or differentiation state. In some embodiments, this prediction is performed by applying gene expression levels of multiple genes from one or more cells in the neuronal progenitor cell population as input to a process configured to predict whether the neuronal progenitor cell population will possess that function, activity, or differentiation state, or whether neurons derived from such cells will possess that function, activity, or differentiation state. In some embodiments, the process includes a machine learning model. In some embodiments, the process (e.g., the machine learning model) is trained using gene expression levels of one or more genes from multiple reference populations of neuronal progenitor cells. In some embodiments, the process (e.g., the machine learning model) is also trained using information about the adaptation of cells in multiple reference populations to a specific characteristic.

[0193] Various machine learning models are suitable for use based on gene expression levels according to the provided methods, and are within the scope of this disclosure. Machine learning models that can be used according to the provided methods include supervised, unsupervised, and semi-supervised machine learning models. In some embodiments, the process (e.g., the machine learning model) is or includes a supervised machine learning model. In some embodiments, the process (e.g., the machine learning model) is or includes an unsupervised machine learning model. In some embodiments, the process (e.g., the machine learning model) is or includes a semi-supervised machine learning model. In some embodiments, the process (e.g., the machine learning model) includes performing one or more data preprocessing techniques. In some embodiments, the process (e.g., the machine learning model) includes performing one or more dimensionality reduction methods. In some examples, machine learning methods include classical machine learning methods such as, but not limited to, Support Vector Machines (SVMs) (e.g., one-class SVMs, linear or radial kernels, etc.), K-Nearest Neighbors (KNN), Isolation Forest, Random Forest, Logistic Regression, AdaBoost classifiers, Extra Tree classifiers, Extreme Gradient Boosting, Gaussian Process classifiers, Gradient Boosting classifiers, Optical Gradient Boosting, Linear Discriminant Analysis, Naive Bayes, Quadratic Discriminant Analysis, Ridge classifiers, or any combination thereof. In some examples, machine learning methods include deep learning methods (e.g., Deep Neural Networks (DNNs)), such as, but not limited to, fully connected networks, Convolutional Neural Networks (CNNs) (e.g., one-class CNNs), Recurrent Neural Networks (RNNs), Transformers, Graph Neural Networks (GNNs), Convolutional Graph Neural Networks (CGNNs), Multilevel Perceptrons (MLPs), or any combination thereof.

[0194] Any suitable method for training machine learning models can be used, including any method described in the following literature: Hastie et al., (2016). The Elements of Statistical Learning; and Abu-Mostafa et al., (2012) Learning from Data (2012). Exemplary machine learning models are also described in Hastie et al., (2016). The Elements of Statistical Learning ; and Abu-Mostafa et al., (2012) Learning from Data .

[0195] In some implementations, classic ML methods include one or more algorithms that learn from existing observations (i.e., known features) to predict outputs. In some implementations, the one or more algorithms perform clustering of the data. In some examples, classic ML algorithms for clustering include K-means clustering, mean-shift clustering, density-based noisy applied spatial clustering (DBSCAN), expectation-maximizing (EM) clustering (e.g., using a Gaussian mixture model (GMM)), aggregated hierarchical clustering, or any combination thereof. In some implementations, the one or more algorithms perform classification of the data. In some examples, classic ML algorithms for classification include logistic regression, Naive Bayes, KNN, random forest, isolated forest, decision tree, gradient boosting, support vector machine (SVM), or any combination thereof. In some examples, SVM includes one-class SVM or multi-class SVM. Classical ML methods may be preferred for small to medium-sized datasets due to their greater interpretability, while deep learning models may be preferred for larger datasets due to their ability to capture complex patterns.

[0196] In some implementations, the process (e.g., a machine learning model) is or includes a regression model. In some implementations, the process (e.g., a machine learning model) is or includes a classification model. In some implementations, the process (e.g., a machine learning model) is or includes a binary classification model. In some implementations, the process (e.g., a machine learning model) is or includes a multi-class classification model.

[0197] In some embodiments, the process (e.g., the machine learning model) is or includes a logistic regression model. In some embodiments, the process (e.g., the machine learning model) is or includes a linear regression model. In some embodiments, the process (e.g., the machine learning model) is or includes a multiple linear regression model. In some embodiments, the process (e.g., the machine learning model) is or includes a multinomial regression model. In some embodiments, the process (e.g., the machine learning model) is or includes a quantile regression model. In some embodiments, the process (e.g., the machine learning model) is or includes a principal component regression model. In some embodiments, the process (e.g., the machine learning model) is or includes a partial minimum regression model. In some embodiments, the process (e.g., the machine learning model) is or includes a support vector regression model. In some embodiments, the process (e.g., the machine learning model) is or includes an ordinal regression model. In some embodiments, the process (e.g., the machine learning model) is or includes a Poisson regression model. In some embodiments, the process (e.g., the machine learning model) is or includes a negative binomial regression model. In some embodiments, the process (e.g., the machine learning model) is or includes a quasi-Poisson regression model. In some embodiments, the process (e.g., the machine learning model) is or includes a Linear Discriminant Analysis (LDA) model. In some embodiments, the process (e.g., the machine learning model) is or includes a Naive Bayes classifier. In some embodiments, the process (e.g., the machine learning model) is or includes a perceptron. In some embodiments, the process (e.g., the machine learning model) is or includes a Support Vector Machine (SVM). In some embodiments, the process (e.g., the machine learning model) is or includes a quadratic classifier. In some embodiments, the process (e.g., the machine learning model) is or includes a decision tree. In some embodiments, the process (e.g., the machine learning model) is or includes a random forest. In some embodiments, the process (e.g., the machine learning model) is or includes a neural network. In some embodiments, the process (e.g., the machine learning model) is or includes an ensemble model having any of the aforementioned models.

[0198] In some embodiments, the process (e.g., the machine learning model) is or includes a penalized machine learning model. A penalized machine learning model is a model that regularizes or constrains coefficient estimates to zero. In some embodiments, the process (e.g., the machine learning model) is or includes a ridge regression model. In some embodiments, the process (e.g., the machine learning model) is or includes a lasso regression model. In some embodiments, the process (e.g., the machine learning model) is or includes a resilient network regression model. In some embodiments, the process (e.g., the machine learning model) is or includes a lasso logistic regression model.

[0199] In some embodiments, the process (e.g., the machine learning model) is or includes a clustering method. In some embodiments, the process (e.g., the machine learning model) is or includes a connectivity-based clustering method. In some embodiments, the process (e.g., the machine learning model) is or includes hierarchical clustering. In some embodiments, the process (e.g., the machine learning model) is or includes a centroid-based clustering method. In some embodiments, the process (e.g., the machine learning model) is or includes k-means clustering. In some embodiments, the process (e.g., the machine learning model) is or includes a distribution-based clustering method. In some embodiments, the process (e.g., the machine learning model) is or includes Gaussian mixture modeling. In some embodiments, the process (e.g., the machine learning model) is or includes a density-based clustering method. In some embodiments, the process (e.g., the machine learning model) is or includes DBSCAN. In some embodiments, the process (e.g., the machine learning model) is or includes OPTICS. In some embodiments, the process (e.g., the machine learning model) is or includes a grid-based clustering method. In some embodiments, the process (e.g., the machine learning model) is or includes STING. In some implementations, the process (e.g., a machine learning model) is or includes CLIQUE.

[0200] In some embodiments, the process (e.g., the machine learning model) is or includes factor analysis. In some embodiments, the process (e.g., the machine learning model) is or includes network component analysis. In some embodiments, the process (e.g., the machine learning model) is or includes linear discriminant analysis. In some embodiments, the process (e.g., the machine learning model) is or includes independent component analysis (ICA). In some embodiments, the process (e.g., the machine learning model) is or includes principal component analysis (PCA). In some embodiments, the process (e.g., the machine learning model) is or includes sparse PCA. In some embodiments, the process (e.g., the machine learning model) is or includes robust PCA.

[0201] In some implementations, the process (e.g., the machine learning model) is or includes nonnegative matrix factorization (NMF). In some implementations, the process (e.g., the machine learning model) is or includes conventional NMF. In some implementations, the process (e.g., the machine learning model) is or includes discriminative NMF. In some implementations, the process (e.g., the machine learning model) is or includes regularized NMF. In some implementations, the process (e.g., the machine learning model) is or includes graph-regularized NMF. In some implementations, the process (e.g., the machine learning model) is or includes bootstrap sparse NMF.

[0202] In some embodiments, the process (e.g., the machine learning model) is or includes kernel PCA. In some embodiments, the process (e.g., the machine learning model) is or includes generalized discriminant analysis (GDA). In some embodiments, the process (e.g., the machine learning model) is or includes an autoencoder. In some embodiments, the process (e.g., the machine learning model) is or includes T-distributed random neighborhood embedding (t-SNE). In some embodiments, the process (e.g., the machine learning model) is or includes manifold learning techniques. In some embodiments, the process (e.g., the machine learning model) is or includes isometric mapping. In some embodiments, the process (e.g., the machine learning model) is or includes locally linear embedding (LLE). In some embodiments, the process (e.g., the machine learning model) is or includes Hessian LLE. In some embodiments, the process (e.g., the machine learning model) is or includes Laplacian eigenmaps. In some embodiments, the process (e.g., the machine learning model) is or includes graph-based kernel PCA. In some embodiments, the process (e.g., the machine learning model) is or includes Unified Manifold Approximation and Projection (UMAP).

[0203] In some implementations, when combining methods for classifying the differentiation states of neurons by predicting their differentiation states, the machine learning models for the first and second reference datasets are of the same type, for example, both are logistic regression models. In some implementations, the machine learning models for the first and second reference datasets are of different types, for example, one is a logistic regression model and the other is a support vector machine classifier. Similarly, the first and second machine learning models trained according to any of the provided methods can be the same or different types of machine learning models.

[0204] In some implementations, the machine learning model includes an ensemble model, which comprises any combination of any of the aforementioned models. Ensemble model techniques, such as model stacking and boosting, can improve prediction accuracy by combining the strengths of different models, particularly in the context of heterogeneous gene expression data.

[0205] In some implementations, these methods include selecting a population of neuronal progenitor cells with desired predictable properties based on the output of a process (e.g., a machine learning model).

[0206] computing devices This disclosure provides a computer system programmed to implement the methods of this disclosure.

[0207] In some embodiments, the present invention provides a computing device configured to predict the transplant viability potential of neuronal progenitor cells when a population of neuronal progenitor cells is implanted into a brain region, the computing device comprising: (a) a processor; and (b) a memory including instructions executable by the processor, the instructions being configured to perform the following steps: (i) receiving a test sample comprising gene expression data of one or more genes (G genes) in the neuronal progenitor cell population associated with the predicted transplant viability potential, wherein the one or more G genes are selected The following groups are free: AC000120.3, KRT77, TTR, PRR16, MEGF10, PDE3A, GDPD2, CMTM8, APOA1, CMTM7, CDHR3, CORIN, VTN, CPNE8, EFEMP1, CD47, SPARC, JAM2, CDO1, PLXDC2, DYNLL2, ITGA3, RPS6KL1, CHRNB2, SULT4A1, PTPN3, LZTS1, RUNX1T1, TMEM 145, EPHA10, CARMIL3, MANEAL, TMEM176B, MPP3, DRAXIN, ADDRB1, KIF26A, CELF5, CNTN2, ASPHD1, SVOP, ANGPT2, SLC22A15, SRRM3, GRIN2D, DACH2, CHST1, GRIN1, LHX5, and NOS2; (ii) determining the gene expression level of each of the one or more G genes based on the test sample; (iii) comparing the determined gene expression level of each of the one or more G genes in the test sample with a reference plot for each corresponding G gene, wherein each reference plot correlates the gene expression level of the G gene with graft size data obtained from a training set including one or more reference samples; and (iv) generating a predictive assessment of the transplant survival potential of a neuronal progenitor cell population by predicting the neuronal transplant survival ability of neuronal progenitor cells in the test sample by correlating the determined gene expression level of the one or more G genes in the test sample with reference plot data.

[0208] In some embodiments, the present invention provides a computing device configured to predict whether neurons differentiating from a population of neuronal progenitor cells will produce dopamine. The computing device includes: (a) a processor; and (b) a memory including instructions executable by the processor, the instructions being configured to perform the following steps: (i) receiving a test sample comprising gene expression data of one or more genes (D genes) in the neuronal progenitor cell population associated with predicted dopamine production potential, wherein the D genes are selected from the group consisting of: CNTNAP5, KLHL1, NHLH2, GREM2, BRINP2, GRIN3A, LRRC4C, IRX3. , CPNE4, PTPN3, PMEL, PCDH20, LRRC37A2, TMEM246, B3GALNT1, ZHX1, BCAS4, SLC25A37, GRINA, MID1, FRMD4A, PAR P10, WHAMMP2, EYA1, CORO2B, WHAMMP3, B3GALT5, GPR35, ABCD2, ITIH3, AC107464.1, CAMK2N1, CAMK2A, PRPS1, GO LGA6L10, AMOT, SULT1A1, CD83, SPON1, FRMPD3, AC096570.1, TCAF2, GOLGA8M, VWA5B2, CA8, AC017050.1, KRT77, AP000350.6, LINC02751, and ARHGAP5-AS1; (ii) determining the gene expression level of each of the one or more D genes based on the test sample; (iii) comparing the determined gene expression level of each of the one or more D genes in the test sample with a reference plot for each corresponding D gene, wherein each reference plot correlates the gene expression level of the D gene with the dopamine production level obtained from a training set including one or more reference samples; and (iv) generating a predictive assessment of the dopamine production potential of derived neurons by predicting the dopamine production capacity of neurons derived from neuronal progenitor cells in the test sample by correlating the determined gene expression level of the one or more D genes with reference plot data.

[0209] Figure 4An example of a computer system is shown, which is programmed or otherwise configured to perform the methods described herein for characterizing populations of neuronal progenitor cells. The computer system includes a central processing unit (CPU, also referred to herein as a “processor” and “computer processor”) 410, which may be a single-core or multi-core processor, or multiple processors for parallel processing. The computer system also includes memory or memory location 440 (e.g., random access memory, read-only memory, flash memory), a communication interface 460 for communicating with one or more other systems (e.g., a network adapter), and peripheral devices 430 (such as cache, other memory, data storage devices, and / or electronic display adapters). Memory 440, interface 460, and peripheral devices 430 communicate with CPU 410 via a communication bus (solid line) (such as a motherboard). Storage unit 470 may be a data storage unit (or data repository) for storing data. The computer system may be operatively coupled to a computer network (“network”) 504 by means of communication interface 460. Figure 5 Network 504 may be the Internet, the Internet and / or an extranet, or an intranet and / or extranet communicating with the Internet. In some cases, network 504 is a telecommunications and / or data network. Network 504 may include one or more computer servers that enable distributed computing, such as cloud computing. In some cases, network 504 may enable a peer-to-peer network, which allows devices coupled to the computer system to act as clients or servers.

[0210] CPU 410 can execute a series of machine-readable instructions, which may be embodied in a program or software. The instructions may be stored in a memory location, such as memory 440. The instructions may be directed to CPU 410, and subsequently, these instructions may program or otherwise configure CPU 410 to implement the methods of this disclosure. Examples of operations performed by CPU 410 may include fetching, decoding, executing, and writing back.

[0211] The CPU 410 can be part of a circuit (such as an integrated circuit). One or more other components of the system may be included in the circuit. In some cases, the circuit is an application-specific integrated circuit (ASIC).

[0212] A graphics processing unit (GPU) is a dedicated processing unit, electronic circuit, module, or computer chip that accelerates many applications and typically exists as a standalone video card, embedded in a motherboard, or as an integrated graphics card on a CPU. Similarly, chip modules capable of performing machine learning predictions (sometimes referred to as inference) are known. Such chips include, for example, language processing units (LPUs), cloud tensor processing units (TPUs), neural engines, AI coprocessors, AI accelerators, and neural processing units (NPUs). In some implementations, GPUs or other chip modules perform at least some of the functions that would otherwise be performed by a CPU.

[0213] Storage unit 470 may store files, such as drivers, libraries, and saved programs. Storage unit may store user data, such as user preferences and user programs. In some cases, the computer system may include one or more additional data storage units 470 located outside the computer system, such as on a remote server communicating with the computer system via an intranet or the Internet. In some embodiments, the computing device includes memory that includes a provided gene expression level dataset. In some embodiments, the dataset is a reference dataset from one or more reference populations of neuronal progenitor cells. In some embodiments, the dataset is a G gene expression level dataset from one or more reference neuronal progenitor cell populations. In some embodiments, the dataset is a D gene expression level dataset from one or more reference neuronal progenitor cell populations. In some embodiments, the memory also includes one or more additional reference datasets. In some embodiments, the one or more additional reference datasets include either a first reference dataset or a second reference dataset as described herein. In some embodiments, the memory also includes a control dataset as described herein.

[0214] A computer system can communicate with one or more remote computer systems via a network 504. For example, the computer system can communicate with a remote computer system belonging to a user (e.g., a laboratory technician preparing neuronal progenitor cells for therapeutic purposes). Examples of remote computer systems include personal computers (e.g., portable PCs), all-in-one or tablet PCs (e.g., Apple...). ® iPad, Samsung ® Galaxy Tab), telephone, smartphone (e.g., Apple) ® iPhone, Android-enabled devices, Blackberry ® (or personal digital assistant). Users may access the computer system via a network 504 error.

[0215] The methods described herein can be implemented by machine-executable code (e.g., a computer processor) stored at an electronic storage location in a computer system, such as, for example, stored in memory 440 or electronic storage unit 470. The machine-executable or machine-readable code may be provided in the form of software. During use, the code can be executed by processor 410. In some cases, the code can be retrieved from storage unit 470 and stored on memory 440 for access by processor 410. In some cases, the electronic storage unit may be excluded, and the machine-executable instructions are stored on memory 440.

[0216] The code can be pre-compiled and configured for use with machines having processors suitable for executing the code, or it can be compiled during runtime. The code is supplied in a programming language, which can be selected to enable the code to be executed either pre-compiled or out-of-bounds.

[0217] Aspects of the systems and methods provided herein (such as computer systems) can be embodied in programming. These various aspects of the technology can be considered as “products” or “articles of art” typically in the form of machine (or processor) executable code and / or associated data carried or embodied in a type of machine-readable medium. The machine-executable code can be stored on electronic storage units, such as memory (e.g., read-only memory, random access memory, flash memory) or hard disk 470. The “storage” medium 470 can include any or all tangible memory of a computer, processor, etc., or associated modules of a computer, processor, etc., such as various semiconductor memories, tape drives, disk drives, etc., which can provide non-transitory storage for software programming at any time. All or part of the software can sometimes communicate via the Internet or various other telecommunications networks. For example, such communication enables the loading of software from one computer or processor to another computer or processor, for example, from a management server or host computer to a computer platform of an application server. Therefore, another type of medium that can carry software elements includes light waves, radio waves, and electromagnetic waves, such as those used over wired and fiber optic fixed networks and over various air links across physical interfaces between local devices. Physical elements carrying such waves (such as wired or wireless links, optical links, etc.) can also be considered as media carrying software. As used herein, unless limited to non-transitory tangible "storage" media, terms such as "computer or machine-readable medium" refer to any medium involved in providing instructions to a processor for execution.

[0218] Therefore, machine-readable media (such as computer-executable code) can take many forms, including but not limited to tangible storage media, carrier media, or physical transmission media. Non-volatile storage media include, for example, optical discs or disks, such as any storage device in any computer, such as a database that can be used to implement the figures shown. Volatile storage media include dynamic memory, such as the main memory of such a computer platform. Tangible transmission media include coaxial cables; copper wires and optical fibers, including conductors that form a bus within a computer system. Carrier transmission media can take the form of electrical or electromagnetic signals or sound or light waves (such as those generated during radio frequency (RF) and infrared (IR) data communications). Therefore, common forms of computer-readable media include, for example: floppy disks, floppy disks, hard disks, magnetic tapes, any other magnetic media, CD-ROMs, DVDs or DVD-ROMs, any other optical media, punched cardstock, any other physical storage media with a perforated pattern, RAM, ROM, PROM and EPROM, FLASH-EPROM, any other memory chip or cartridge, a carrier wave for transmitting data or instructions, a cable or link for transmitting such a carrier wave, or any other medium from which a computer can read programming code and / or data. Many of these forms of computer-readable media may involve carrying one or more sequences of one or more instructions to a processor for execution.

[0219] The computer system may include or communicate with an electronic display, the electronic display including a user interface (UI) 430 for providing an evaluation of, for example, a population of neuronal progenitor cells. Examples of UIs include, but are not limited to, graphical user interfaces (GUIs) and web-based user interfaces.

[0220] The methods and systems disclosed herein can be implemented by one or more algorithms. The algorithms can be implemented in software when executed by the central processing unit 410. The algorithms can automatically characterize, for example, a population of neuronal progenitor cells with respect to: the likelihood of successful transplantation and survival of the neuronal progenitor cell population when implanted into the brain of a subject; the likelihood of the neuronal progenitor cell population generating dopamine-producing neurons; and / or determining whether the cells are in the desired differentiation state.

[0221] Compositions and formulations In embodiments of the provided methods, neuronal progenitor cells identified by the methods provided herein may be harvested or collected, for example for formulation and use as therapeutic agents, such as for use in cell therapies for treating neurodegenerative diseases. In some embodiments, a dose of cells including neuronal progenitor cells (e.g., identified DA neuronal progenitor cells) is provided as a composition or formulation, such as a pharmaceutical composition or formulation. Such compositions may be used according to the provided methods, such as for the prevention or treatment of neurodegenerative conditions, including Parkinson's disease.

[0222] In some embodiments, the provided method further includes selecting cells predicted to produce cells with the ability to transplant successfully and / or produce dopamine. In some embodiments, the selected population is harvested according to any of the methods described herein. In some embodiments, harvesting takes place between approximately day 13 and approximately day 20. In some embodiments, harvesting takes place on approximately day 16 or later. In some embodiments, the selected population is harvested at approximately day 16 or later of culture. In some embodiments, harvesting takes place between approximately day 18 and approximately day 24 of culture. In some embodiments, the selected population is harvested at approximately day 18 to approximately day 23 of culture. In some embodiments, harvesting takes place on day 18, day 19, day 20, day 21, day 22, or day 23, or on approximately day 18, approximately day 19, approximately day 20, approximately day 21, approximately day 22, or approximately day 23. In some embodiments, the selected population is harvested on day 18, 19, 20, 21, 22, or 23 of cultivation, or approximately day 18, 19, 20, 21, 22, or 23 of cultivation. In some embodiments, harvesting is carried out on day 20 or approximately day 20. In some embodiments, the selected population is harvested on day 20 of cultivation or approximately day 20 of cultivation.

[0223] This document also provides compositions comprising cell populations generated or selected according to the provided methods. This document also provides methods for treating subjects suffering from neurodegenerative diseases, wherein the subjects are treated by implantation of any of the provided compositions.

[0224] The provided embodiments relate to methods for generating differentiated neuronal progenitor cells suitable for administration to a subject to treat a neurodegenerative disease. In certain embodiments, the methods provided herein enhance the ability of therapeutic cell compositions to generate differentiated neuronal progenitor cells that produce cells more likely to survive transplantation (e.g., also achieving innervation) and / or be capable of producing dopamine in brain regions of a subject suffering from a neurodegenerative disease. In some specific embodiments, the neurodegenerative disease is Parkinson's disease (PD). In some embodiments, the provided methods address problems related to the characteristics of Parkinson's disease (PD), including selective degeneration of midbrain dopamine (mDA) neurons in the patient's brain. Because PD symptoms are primarily due to the selective loss of DA neurons in the ventral substantia nigra of the midbrain, PD is considered suitable for cell replacement therapy strategies.

[0225] In some embodiments, the provided therapeutic composition is a pharmaceutical composition containing a pharmaceutically acceptable carrier. In some embodiments, a cell dose comprising cells classified by any of the methods disclosed herein is provided in the form of a composition or formulation, such as a pharmaceutical composition or formulation. Such compositions may be used, for example for the prevention or treatment of diseases, symptoms, and conditions, such as neurodegenerative diseases, depending on the provided methods, articles, and / or compositions provided.

[0226] In some embodiments, the provided method further includes preparing the harvested cells with a cryoprotectant. In some embodiments, the harvested cells are prepared according to any of the methods described herein. In some embodiments, the provided method further includes cryopreserving the prepared cells. In some embodiments, the prepared cells are cryopreserved according to any of the methods described herein. In some embodiments, cryopreservation includes controlled-rate freezing.

[0227] In some embodiments, the provided method further includes preparing the harvested cells with a cryoprotectant. In some embodiments, the harvested cells are prepared according to any of the methods described herein. In some embodiments, the provided method further includes cryopreserving the prepared cells. In some embodiments, the prepared cells are cryopreserved according to any of the methods described herein. In some embodiments, cryopreservation includes controlled-rate freezing.

[0228] In some cases, processing cells in one or more steps to manufacture, generate, or produce cell therapy and / or differentiated cells may include cell formulations, such as formulations of differentiated cells produced by the methods. In some cases, cells may be formulated in amounts intended for dosage administration, such as for single unit dose administration or multiple dose administration.

[0229] In some embodiments, one or more compositions are formulated to differentiate cells. In particular embodiments, one or more compositions are formulated to differentiate cells after the production of one or more compositions. In some embodiments, the one or more compositions have been previously cryopreserved and stored, and thawed prior to application.

[0230] Products and reagent kits In some embodiments, this document provides kits for performing any of the methods provided. In some embodiments, the kit includes one or more primer sets and optional probes for amplifying a gene (e.g., a G gene or a D gene). In some embodiments, each primer set and probe is specific to the gene to be amplified.

[0231] In some implementations, this document provides a kit comprising one or more pairs of oligonucleotide primers for predicting transplant survival, wherein each pair of oligonucleotide primers is specific for genes selected from the group consisting of: AC000120.3, KRT77, TTR, PRR16, MEGF10, PDE3A, GDPD2, CMTM8, APOA1, CMTM7, CDHR3, CORIN, VTN, CPNE8, EFEMP1, CD47, SPARC, JAM2, CDO1, PLXDC2, DYNL L2, ITGA3, RPS6KL1, CHRNB2, SULT4A1, PTPN3, LZTS1, RUNX1T1, TMEM145, EPHA10, CARMIL3, MANEAL, TMEM176B, MPP3, DRAXIN, ADGRB1, KIF26A, CELF5, CNTN2, ASPHD1, SVOP, ANGPT2, SLC22A15, SRRM3, GRIN2D, DACH2, CHST1, GRIN1, LHX5, and NOS2. In some embodiments, the kit includes two pairs of oligonucleotide primers specific to any two of the above genes. In some embodiments, the kit includes three pairs of oligonucleotide primers specific to any two of the above genes. In some embodiments, the kit includes four pairs of oligonucleotide primers specific to any two of the above genes. In some embodiments, the kit includes five pairs of oligonucleotide primers specific to any two of the above genes. In some embodiments, the kit includes six pairs of oligonucleotide primers specific for any two of the genes described above. In any such embodiment, the kit may also include oligonucleotide probes for each primer pair. In some embodiments, the kit is used to predict the transplant viability of neurons derived from neuronal progenitor cells.

[0232] In some embodiments, this document provides a kit comprising three primer pairs: a first oligonucleotide primer pair suitable for amplifying a first gene, a second oligonucleotide primer pair suitable for amplifying a second gene, and a third oligonucleotide primer pair suitable for amplifying a third gene. In some embodiments, the first, second, and third genes are each selected from the group consisting of: TTR, PRR16, CMTM8, APOA1, CD47, CD01, KIR26A, and CNTN2. In some embodiments, the first gene is TTR, the second gene is PRR16, and the third gene is CD47. In any such embodiment, the kit may also include oligonucleotide probes for each primer pair. In some embodiments, the kit is used to predict the transplant viability of neurons derived from neuronal progenitor cells.

[0233] In some implementations, this document provides kits comprising one or more pairs of oligonucleotide primers for predicting dopamine production, wherein each pair of oligonucleotide primers is specific for genes selected from the group consisting of: CNTNAP5, KLHL1, NHLH2, GREM2, BRINP2, GRIN3A, LRRC4C, IRX3, CPNE4, PTPN3, PMEL, PCDH20, LRRC37A2, TMEM246, B3GALNT1, ZHX1, BCAS4, SLC25A37, GRINA, MID1, FRMD4A, PAR P10, WHAMMP2, EYA1, CORO2B, WHAMMP3, B3GALT5, GPR35, ABCD2, ITIH3, AC107464.1, CAMK2N1, CAMK2A, PRPS1, GOLGA6L10, AMOT, SULT1A1, CD83, SPON1, FRMPD3, AC096570.1, TCAF2, GOLGA8M, VWA5B2, CA8, AC017050.1, KRT77, AP000350.6, LINC02751, and ARHGAP5-AS1. In some embodiments, the kit includes two pairs of oligonucleotide primers specific to any two of the above genes. In some embodiments, the kit includes three pairs of oligonucleotide primers specific to any two of the above genes. In some embodiments, the kit includes four pairs of oligonucleotide primers specific to any two of the above genes. In some embodiments, the kit includes five pairs of oligonucleotide primers specific to any two of the genes described above. In some embodiments, the kit includes six pairs of oligonucleotide primers specific to any two of the genes described above. In any such embodiment, the kit may also include oligonucleotide probes for each primer pair. In some embodiments, the kit is used to predict dopamine production in neurons derived from neuronal progenitor cells.

[0234] In some embodiments, this document provides a kit comprising three primer pairs: a first oligonucleotide primer pair suitable for amplifying a first gene, a second oligonucleotide primer pair suitable for amplifying a second gene, and a third oligonucleotide primer pair suitable for amplifying a third gene. In some embodiments, the first, second, and third genes are each selected from the group consisting of: CNTNAP5, NHLH2, GREM2, PMEL, PCDH20, LRRC37A2, SLC25A37, MID1, EYA1, B3GALT5, GPR35, AC107464.1, CAMK2N1, CAMK2A, GOLGA6L10, FRMPD3, VWA5B2, AC017050.1, and LINC02751. In some embodiments, the first gene is B3GALT5, the second gene is GREM2, and the third gene is FRMPD3. In any such embodiment, the kit may also include oligonucleotide probes for each primer pair. In some implementations, the kit is used to predict the transplant survival potential of neurons derived from neuronal progenitor cells.

[0235] In some embodiments, each oligonucleotide primer in the oligonucleotide primers is operatively linked to a detectable marker. In some embodiments, the detectable marker is a fluorescent label.

[0236] In some embodiments, the kit also includes probes for each pair of oligonucleotide primers. In some embodiments, the probes include a reporter dye and a quencher. In some embodiments, the reporter dye for each probe in the kit is different. In some embodiments, the reporter dye is a fluorescent label.

[0237] In some embodiments, this document also provides articles comprising any of the provided therapeutic compositions. In some embodiments, this document also provides kits comprising (i) any of the provided therapeutic compositions and (ii) instructions for administering the therapeutic composition to a subject.

[0238] In some embodiments, the article or kit includes one or more containers (typically multiple containers), packaging materials, and labels or instructions for use on or associated with one or more containers and / or packaging. In some embodiments, the instructions for use provide guidance or specify methods for assessing whether a subject is likely or suspected of being able to produce a response and / or the extent or level of a response prior to receiving cell therapy. In some aspects, the article may contain a dose of differentiated cells or a composition of differentiated cells.

[0239] The articles of manufacture provided herein contain packaging materials. Packaging materials used to package the provided materials are well known to those skilled in the art. See, for example, U.S. Patents 5,323,907, 5,052,558, and 5,033,252, each of which is incorporated herein in its entirety. Examples of packaging materials include, but are not limited to, blister packs, bottles, tubes, inhalers, pumps, bags, vials, containers, syringes, disposable laboratory supplies such as pipette tips and / or plastic sheets, or bottles. Articles of manufacture or kits may include a device to facilitate the dispensing of materials or to facilitate use in a high-throughput or large-scale manner, such as to facilitate use in robotic equipment. Typically, the packaging does not react with the therapeutic composition contained therein.

[0240] In some embodiments, these compositions are individually packaged. In some embodiments, each container may have a single compartment. In some embodiments, other components of the article or kit are packaged separately or together in a single compartment.

[0241] Treatment In some embodiments, this document provides methods for treating a subject with a disease or condition requiring treatment using any of the provided therapeutic compositions. In some embodiments, the provided methods include implanting a population of neuronal progenitors having one or more desired characteristics into the subject. In some embodiments, the subject suffers from a neurodegenerative disease. In some embodiments, the neurodegenerative disease includes the loss of dopamine neurons in the brain. In some embodiments, the subject has lost dopamine neurons in the substantia nigra (SN). In some embodiments, the subject has lost dopamine neurons in the pars compacta (SNc). In some embodiments, the subject exhibits rigidity, bradykinesia, impaired postural reflexes, resting tremor, or a combination thereof. In some embodiments, the subject exhibits abnormal [18F]-L-DOPA PET scans. In some embodiments, the subject exhibits [18F]-DG-PET evidence of Parkinson's Disease-Related Patterns (PDRP).

[0242] In some embodiments, the neurodegenerative disease is Parkinson's syndrome. In some embodiments, the neurodegenerative disease is Parkinson's disease. In some embodiments, the neurodegenerative disease is idiopathic Parkinson's disease. In some embodiments, the neurodegenerative disease is a familial form of Parkinson's disease. In some embodiments, the subject has mild Parkinson's disease. In some embodiments, the subject has a Movement Disorder Association-Unified Parkinson's Disease Rating Scale (MDS-UPDRS) motor score of 32 or less. In some embodiments, the subject has moderate or advanced Parkinson's disease. In some embodiments, the subject has mild Parkinson's disease. In some embodiments, the subject has an MDS-UPDRS motor score between 33 and 60.

[0243] In some embodiments, the dose of cells is administered to the striatum of the subject. In some embodiments, the dose of cells is administered to one hemisphere of the striatum of the subject. In some embodiments, the dose of cells is administered to both hemispheres of the striatum of the subject. In some embodiments, the cell dose is from exactly or about 1 million cells per hemisphere to exactly or about 30 million cells per hemisphere. In some embodiments, the cell dose is from exactly or about 5 million cells per hemisphere to exactly or about 20 million cells per hemisphere. In some embodiments, the cell dose is from exactly or about 10 million cells per hemisphere to exactly or about 15 million cells per hemisphere.

[0244] Example The following examples are included for illustrative purposes only and are not intended to limit the scope of the invention.

[0245] Two types of models are described in the following examples. The first is a PCA-based model, and the second is a gene-based model. The PCA-based model can be used to predict graft size based on the bulk gene expression levels of dopaminergic neuronal progenitor cells (DANPCs), or to predict the amount of dopamine released by DANPCs in vitro during extended culture upon the addition of KCl. Similarly, the gene-based model can also be used to predict graft size based on the bulk gene expression levels of DANPCs, or to predict the amount of dopamine released by DANPCs in vitro during extended culture upon the addition of KCl. The following examples illustrate how these two different models can be used to predict graft size or the amount of dopamine released.

[0246] Example 1—PCA-based model for predicting graft size A machine learning method was developed to predict DANPC transplantation outcomes based on DANPC gene expression levels. To this end, bulk gene expression levels of a subset of cells from a human DANPC population differentiated from in vitro pluripotent stem cells (iPSCs) were determined. The remaining cells from the same human DANPC population were stereotactically injected unilaterally or bilaterally into the striatum of rats. The injected DANPC was then quantified using image processing software to obtain various features related to DANPC transplantation outcomes. Among these transplantation outcomes, the graft size of the injected DANPC was determined. Thus, population-specific bulk gene expression levels and population-specific graft sizes were obtained from a given human DANPC population. For DANPC populations derived from multiple human subjects, population-specific bulk gene expression levels and population-specific graft sizes were also obtained. The PCA-based model described in this embodiment uses DANPC gene expression levels to predict the graft size of the injected DANPC. Figure 6 A schematic workflow for a PCA-based model to predict graft size using bulk gene expression data from DANPC is described. Notably, the method described herein is not limited to predicting graft size in transplanted DANPC, but can be generalized to a) predicting any kind of transplant outcome, such as transplant outcomes that are not graft size-related, and b) other cell types that are not DANPC.

[0247] Training and testing data used to train and test the machine learning model included bulk RNA sequencing (RNA-seq) data of DANPCs. Bulk gene expression levels were then correlated with the graft size of the corresponding DANPC. The strength of the correlation between graft size and RNA-seq expression of the corresponding DANPC was quantified using metrics such as the Pearson correlation coefficient, based on the Weighted Correlation Network Analysis (WGCNA) package in the R programming language.

[0248] A. The data is divided into training and testing parts based on the donor. The Leave-One-Out (LOO) method divides the resulting data on the correlation between batch RNAseq data and corresponding DANPC graft sizes into training and test data portions. This process iterates the same number of times as the number of donors represented in the correlation data, for example, 7 iterations. The LOO method is a cross-validation approach. The purpose of cross-validation is to control for the undesirable possibility that the training data portion may not represent the variable of interest used for prediction, and therefore, the model may fail to generalize because it is trained on unusual data. The training data portion can randomly include very unusual data compositions, and training the model based on such unusual compositions may cause the model to fail to generalize across all or most instances of the variable of interest used for prediction. Cross-validation is used to limit this artifact. Cross-validation requires generating many pairs of training and test data portions. In the case of LOO cross-validation, a single data point is reserved as test data, and the remaining data is treated as training data. In this embodiment, the DANPC originates from 7 subjects, so 7 iterations of LOO are used for cross-validation: 1 subject (e.g., subject A) for test data, and the remaining 6 subjects (subjects B to G) for training data; 1 subject (e.g., subject B) for test data, and the remaining 6 subjects (subjects A, C to G) for training data; ...; 1 subject (e.g., subject G) for test data, and the remaining 6 subjects (subjects A to F) for training data. LOO cross-validation is useful when the sample size of independent and separately distributed (IID) random variables is small. This embodiment treats the relevant sample size as the number of subjects, which differs from many other methods that treat the relevant sample size as the number of cells to be transplanted (e.g., DANPC). Although using the number of subjects as the relevant sample size generally results in a smaller sample size than using the number of cells, choosing the number of subjects as the relevant sample size during cross-validation can improve the predictive ability of the machine learning model. However, some methods choose to use the number of cells as the sample size because doing so often enlarges the sample size, and experimenters may consider a larger sample size unconditionally advantageous. However, using the number of cells to be transplanted as the relevant sample size can often lead to poorer predictive power in machine learning models because, in cases where the transplanted cells originate from a common donor, cells from the common donor are not true IIDs—cells from the common donor may have more in common than cells from another donor. Furthermore, by treating the relevant sample size as the number of subjects rather than the number of cells to be transplanted, splitting the data into training and testing portions during cross-validation does not result in cells from a given donor being present in both the training and testing portions.This model design choice is particularly important when using RNAseq data from donor cells, because most unique expression features in donor cell RNAseq data indicate traits specific to the donor subject, rather than more general variables of interest common to most donors, such as cell transplantation outcomes, for example, graft size. For the reasons stated above, the method described in this embodiment chooses to use the number of subjects rather than the number of cells used for transplantation as the relevant sample size for cross-validation, even if doing so may result in fewer iterations for cross-validation. Therefore, the cross-validation method used in this embodiment is limited to cross-validation methods compatible with small sample sizes, and thus LOO is used.

[0249] B. Perform on five randomly sampled genes PCA For each of the seven iterations, a Loo operation was performed on the correlation data (i.e., correlation data regarding the correlation between a given gene expression level from DANPC bulk RNAseq data and the corresponding DANPC graft size). The correlation data in the training subset was sorted, and the top 2000 correlation values ​​were selected from the training subset of each iteration. For each training subset, 2000 genes corresponding to the 2000 correlation values ​​were selected, and genes common to all seven iterations were identified. Then, all genes common to all seven iterations were filtered for expression levels greater than 0 log counts per million (log CPM). Four hundred and fifteen genes were identified that were common to the training subsets of all seven Loo iterations and remained after the expression filtering. From these 415 identified genes, five were randomly selected, and principal component analysis (PCA) was performed on these five genes. Five genes were randomly selected from 415 identified genes, and PCA was performed on these five randomly selected genes. This process was repeated one million times, resulting in one million PCA results (e.g., the percentage of data variance explained by each principal component generated by each PCA). R was determined for each portion of the training data used to generate each of these one million PCA results. 2 The R-value (i.e., the ratio of RNAseq levels of the 5 selected genes to the corresponding graft size) 2 The result of one million PCA operations is a dataset consisting of one million rows, where each row includes the corresponding R value. 2 The values, and the percentage of data variance explained by each principal component (PC) generated by each PCA. It is worth noting that running one million PCA sessions represents a computational efficiency improvement compared to the full number of PCA sessions required to study each possible selection of 5 genes from 415 identified genes. The total number of possible combinations of 5 genes drawn from 415 possible genes is... = 100,128,170,583 possible combinations. Therefore, performing PCA on only one million sets of five genes randomly from these 100,128,170,583 possible combinations saves approximately six orders of magnitude in computational efficiency. Furthermore, selecting rarely used genes (e.g., five genes) from a total of 415 possible genes controls for other significant influences that a particular donor subject will have when training machine learning models (e.g., models derived from PCA). If a large number of genes are selected from the total of 415 possible genes, and PCA is then performed on these large numbers of genes, most of the variance captured by PCA will be irrelevant to general transplant outcomes (i.e., transplant outcomes common to or unknown to donor subjects) and will instead be primarily related to traits specific to the donor subject. Therefore, to avoid overfitting PCA to the training data, only a small number of genes (e.g., five genes) are repeatedly selected from the 415 possible genes.

[0250] C. A linear regression model was generated based on PC1 and the known graft size. Each row of the 1 million-row dataset describing the PCA results is used to predict graft size based on the test portion via linear regression. Some results from each of the 1 million PCA operations can be described as a matrix of 22 cell batches × 5 PCs, where only the column vector describing the results of the first PC (i.e., PC1) (i.e., 22 cell batches × PC1) is of interest. More specifically, the elements of the 22 cell batches × PC1 vector are projections of gene expression (e.g., batch RNAseq) data onto PC1 (i.e., the first set of coordinates in PC space). The 22 projections onto PC1 are plotted against the graft sizes of the 22 corresponding cell batches from the training portion, and linear regression is performed, i.e., generating the form: The predictive linear model, where y This is the predicted graft size. Then, the test data portion is fed into a predictive linear model (based on the training data portion) to generate the predicted graft size. Therefore, each PCA operation corresponds to a set of 5 genes from 506 possible genes via a linear relationship, and the resulting predicted graft size. Figure 7 The paper depicts the obtained linear relationships for 10 genes, among which... Figure 7 In each plot, the x-axis represents the gene expression of the gene of interest, and Figure 7 In each plot, the y-axis represents the graft size. It's worth noting that the R-squared, describing the fit of the linear regression to the 22 data points, was also calculated. 2The method described in this embodiment generated 1 million PCA analyses, which provided the basis for 1 million predictive linear regressions (i.e., regressions of data projected onto PC1 against known graft sizes from the training subset), which predicted 1 million graft sizes. For genes 1 through 5, increased gene expression was associated with larger graft sizes, and for genes 6 through 10, increased gene expression was associated with smaller graft sizes.

[0251] D. Model grouping, folding, and sorting Then, the best model was selected from 1 million linear regression models. For this purpose, x-axis values ​​(PC1%) were generated, including the percentage of variance explained by PC1 and R-axis values. 2 The y-axis grid values. PC1% ranges from 0% to 100%, and R... 2 The values ​​range from 0 to 1. The predicted graft size is determined by a) PC1% values ​​between 0% and 10% and b) R values ​​between 0 and 0.1. 2 Up to 1000 linear regression models were examined to provide the basis for the values, and the predicted graft sizes were averaged, with cross-validation statistics such as AUC (area under the receiver-operator curve) calculated. In this way, the values ​​were calculated based on their corresponding PC1% values ​​and R... 2 The system groups up to 1000 linear regression models and folds those models together by averaging their predicted graft sizes and calculating individual cross-validation summation statistics for the folded linear regression models, such as individual accuracy and AUC values ​​(and R-squared values). 2 (Sensitivity, specificity, precision, recall, and F1 score). Averaging the predicted graft size buffers most of the differences seen in the predicted graft size across different models.

[0252] Repeat the process described above, but use different ranges of PC1% and R. 2 The values ​​were used to group up to 1000 linear regression models. Specifically, the PC1% range increased by 2.5 percentage points. Therefore, the second group of up to 1000 linear regression models included a) PC1% ranging from 2.5% to 12.5%, and b) R values ​​remaining between 0 and 0.1. 2 Value. Similarly, collapse at the specified PC1% and R in the following way. 2Up to 1000 linear regression models within the range of values: averaging the graft size predicted by the models, and calculating individual cross-validation statistics, such as individual accuracy and AUC values, as well as the other metrics mentioned above, for all folded models. The group of up to 1000 linear regression models continues to increase in increments of 2.5% along the x-axis of the grid (and within the PC1% and R...). 2 At each distinct group within the range, the predicted graft size is averaged, and the cross-validation statistic is calculated across the entire group, until for this group of up to 1000 linear regression models, the PC1% range becomes at least 90% (assuming a window size of 10%, the window will not move more than 90%, otherwise the effective window size will become less than 10%, and reasonable comparisons against other collapsed models will not be possible). The next group of up to 1000 linear regression models then includes a) a PC1% range between 2.5% and 12.5%, and b) an R-value between 0.025 and 0.125. 2 The values ​​were calculated, and similarly, the predicted graft sizes for up to 1000 models were averaged, and the cross-validation statistic was calculated across the entire group. Based on the PC1% value and R... 2 The iterative process of grouping up to 1000 linear regression models by the range of values ​​continues until the PC1% value is between 0% and 100% and the R-value is between 0 and 1. 2 The entire space of values ​​is filled. As seen above, for each group in the linear regression model, increasing its PC1% value by 2.5 percentage points or its R-value... 2 The value increases by 0.025 (i.e., both the x-axis and y-axis increase by 2.5%). Once the PC1% value and R... 2 If the entire space of values ​​is filled, then up to 1,000 groups of folded linear regression models can be compared via group statistics of interest (e.g., cross-validation statistics).

[0253] It is worth noting that when up to 1000 models are folded, the predicted graft sizes are averaged. Strictly speaking, averaging the predicted graft sizes is not necessary, but it allows them to withstand different forms of central tendency, such as the median. The median-based predicted graft size may be more robust than the mean-based prediction for any outliers present in the data from the cell batch.

[0254] Figure 8 A to Figure 8 H shows the PC1% and R based on up to 1000 linear regression models. 2 The values ​​are used to group these linear regression models. Figure 8 A to Figure 8 Each plot in H is a heatmap drawn relative to the group statistic of interest, where the x-axis is PC1% and the y-axis is the R-squared value of the linear regression model. 2,Right now, Figure 8 A to Figure 8 Each plot in H depicts the PC1%-R described above. 2 Spatial grid. That is, Figure 8 A to Figure 8 H groups the R values ​​of the model in the corresponding order. 2 Heatmaps were plotted using AUC, sensitivity, specificity, precision, F1, accuracy, and kappa value. Figure 8 A to Figure 8 In H, the black outline circle in each plot indicates the model with the highest AUC and highest accuracy (from those with similar PC1% and R). 2 (The values ​​are derived from the folded model group). This embodiment sorts the best group of up to 1000 linear regression models, first by the highest accuracy, and then by the highest AUC.

[0255] Figure 9 A to Figure 9 B shows the predictive properties of a set of PCA-based linear regression models. Figure 9 A describes the relationship between the predicted graft size (based on an arbitrary scale normalized to arbitrary units from 0 to 1) and the measured graft size (also based on an arbitrary scale normalized to arbitrary units from 0 to 1) for the (folded) model that first has the highest AUC and then the highest accuracy. Figure 9 B describes the receiver-operator characteristic curve (ROC curve) and its corresponding AUC for the model with the highest AUC and the highest accuracy.

[0256] Figure 10 The various properties of the model that first has the highest AUC and then the highest accuracy (i.e., the optimal model) are described (to recap, this model is achieved by folding with similar R values). 2 (Based on up to 1000 models with PC1% values). The best model included an accuracy of 0.8636, an AUC of 0.875, and an R-value of 0.7353. 2 It has a sensitivity of 0.9, a specificity of 0.8333, a precision of 0.8182, a recall of 0.9, and an F1 score of 0.8571.

[0257] The highest-ranked group of up to 1000 linear regression models can be selected based on any desired criteria, allowing the identification of a desired number of unique genes from the corresponding 5-gene set of those model groups.

[0258] Example 2—A non-PCA-based model for predicting graft size A machine learning method was developed to predict DANPC transplantation outcomes based on gene expression levels of dopaminergic neuronal progenitor cells (DANPCs). To this end, bulk gene expression levels were determined from a subset of cells in a human DANPC population differentiated from in vitro pluripotent stem cells (iPSCs). The remaining cells from the same human DANPC population were stereotactically injected unilaterally or bilaterally into the striatum of rats. The injected DANPCs were then quantified using image processing software to obtain various features related to DANPC transplantation outcomes. Among these transplantation outcomes, the graft size of the injected DANPCs was determined. Thus, population-specific bulk gene expression levels and population-specific graft sizes were obtained from a given human DANPC population. For DANPC populations derived from multiple human subjects, population-specific bulk gene expression levels and population-specific graft sizes were also obtained. The non-PCA-based model described in this embodiment uses DANPC gene expression levels to predict the graft size of injected DANPCs.

[0259] Training and testing data for the machine learning models included batch RNA sequencing (RNAseq) data from DANPCs. Batch gene expression levels were then modeled relative to the graft size of the corresponding DANPC using the edgeR package in the R programming language. The edgeR package models the relationship between individual gene expression and the response variable "graft size" using an overly discrete Poisson model, such that each gene receives its own estimate from a generalized linear model likelihood ratio test. Based on the overly discrete Poisson model, the top 1000 genes associated with graft size were identified for a given training subset.

[0260] A. The data is divided into training and testing parts based on the donor. The Leave-One-Out (LOO) method splits the resulting data on the association between batch RNAseq data and corresponding DANPC graft sizes into training and test data portions. This process iterates the same number of times as the number of donors represented in the edgeR modeling data, for example, 7 iterations. The LOO method is a cross-validation approach. The purpose of cross-validation is to control for the undesirable possibility that the training data portion may not represent the variable of interest used for prediction, and therefore, the model may fail to generalize because it is trained on unusual data. The training data portion can randomly include very unusual data compositions, and training the model based on such unusual compositions may cause the model to fail to generalize across all or most instances of the variable of interest used for prediction. Cross-validation is used to limit this artifact. Cross-validation requires generating many pairs of training and test data portions. In the case of LOO cross-validation, a single data point is reserved as test data, and the remaining data is treated as training data. In this embodiment, the DANPC originates from 7 subjects, so 7 iterations of LOO are used for cross-validation: one subject (e.g., subject A) is used for the test data, and the remaining six subjects (subjects B through G) are used for the training data; one subject (e.g., subject B) is used for the test data, and the remaining six subjects (subjects A, C through G) are used for the training data; ...; one subject (e.g., subject G) is used for the test data, and the remaining six subjects (subjects A through F) are used for the training data. LOO cross-validation is useful when the sample size of independent and separately distributed (IID) random variables is small. This embodiment treats the relevant sample size as the number of subjects, which differs from many other methods that treat the relevant sample size as the number of cells to be transplanted (e.g., DANPC). Although using the number of subjects as the relevant sample size generally results in a smaller sample size than using the number of cells, choosing the number of subjects as the relevant sample size during cross-validation can improve the predictive ability of the machine learning model. However, some methods choose to use the number of cells as the sample size because doing so often increases the sample size, and experimenters may consider a larger sample size unconditionally advantageous. However, using the number of cells to be transplanted as the relevant sample size can often lead to poorer predictive power in machine learning models because, in cases where the transplanted cells originate from a common donor, cells from the common donor are not true IIDs—cells from the common donor may have more in common than cells from another donor. Furthermore, by treating the relevant sample size as the number of subjects rather than the number of cells to be transplanted, splitting the data into training and testing portions during cross-validation does not result in cells from a given donor being present in both the training and testing portions.This model design choice is particularly important when using RNAseq data from donor cells, because most unique expression features in donor cell RNAseq data indicate traits specific to the donor subject, rather than more general variables of interest common to most donors, such as cell transplantation outcomes, for example, graft size. For the reasons stated above, the method described in this embodiment chooses to use the number of subjects rather than the number of cells used for transplantation as the relevant sample size for cross-validation, even if doing so may result in fewer iterations for cross-validation. Therefore, the cross-validation method used in this embodiment is limited to cross-validation methods compatible with small sample sizes, and thus LOO is used.

[0261] B. Calculations were performed on three randomly sampled genes. Based on edgeR-based modeled gene expression data and their association with graft size, the top 1000 genes associated with graft size were identified for each of the seven training parts. Then, intersecting genes across each of the seven training parts were identified. In addition to gene intersecting, each gene needed to have at least 40 reads mapped to it in at least six training part RNAseq libraries. Based on these criteria, 98 intersecting genes were identified. These 98 genes were then further filtered using 40 reads mapped to a given gene in at least six training part RNAseq libraries, retaining only the more highly expressed genes. After filtering for highly expressed genes, 62 genes remained. Figure 11 A schematic workflow for a non-PCA-based model to predict graft size using bulk gene expression data from DANPC is described.

[0262] Three genes are randomly sampled from the remaining 62 genes. The total number of possible combinations of the three genes drawn from the 62 possible genes is: = 37,820 possible combinations. To achieve greater computational efficiency, only a set of 3 genes was randomly sampled 20,000 times from the 37,820 possibilities, thus improving computational efficiency by nearly four times. Furthermore, selecting rarely used genes (e.g., 3 genes) from the total of 62 possible genes controls for other significant influences that a particular donor subject will have when generating one or more models used to predict graft size. If a large number of genes were selected from the total of 62 possible genes, much of the predictive power based on gene expression data would be irrelevant to general transplant outcomes (i.e., transplant outcomes common to or unknown to donor subjects) and would instead be primarily related to the specific traits of the donor subject. Therefore, to avoid generating models based on overfitting training data, only a small number of genes (e.g., 3 genes) are repeatedly selected from the 62 possible genes.

[0263] C. A linear regression model is generated based on gene expression and known graft size. To generate a model predicting graft size for each of the three randomly sampled gene combinations, a linear regression was performed between the gene expression level of each of the three genes and the graft size of 22 cell batches, in the form of... The linear regression models were used. To recap, the dataset analyzed in this embodiment comprised 22 cell batches from 7 donor subjects, each with corresponding batch RNAseq data including expression data for the first 62 genes. Therefore, each linear regression model was derived from the sample size of the 22 cell batches, where the predictor variable is the batch gene expression of one of three randomly selected genes from a given cell batch, and the response variable is the graft size of the given cell batch.

[0264] Figure 12 A to Figure 12 C describes a linear regression model for each of three randomly selected genes from a list of 62 genes. Figure 12 A to Figure 12 In C, the x-axis of each plot represents the gene expression of one of three randomly selected genes (in this case, TTR). Figure 12 A), CD47 ( Figure 12 B) and PRR16 Figure 12 C), and the y-axis of each plot represents the graft size of the cell batch. Each data point on each plot corresponds to one of the 22 cell batches. For emphasis, each linear regression model in this embodiment is based on data from the training portion. The linear regression model can then be used to predict graft size based on gene expression data from the test portion. For each set of three linear regression models corresponding to three randomly selected sets of genes, the three predicted graft sizes are averaged to generate a single predicted graft size. It is worth noting that the average can be replaced with some other type of central tendency calculation, such as the median, which tends to be more robust than the mean for outliers. The three linear regression models for each set are folded together, and for the folded model, a single accuracy, a single AUC, and other single cross-validation summary statistics are determined.

[0265] Figure 13 A depicts the relationship between measured graft size and predicted graft size obtained using a gene-based model, and Figure 13 B describes the targets based on TTR, CD47, and PRR16 (such as...). Figure 12 A to Figure 12The receiver operating characteristic (ROC) curves and their associated AUC values ​​for the 3-gene model (shown in Figure C). Figure 14 Depicting and Figure 13 A to Figure 13 The folding model described in section B is associated with several summary statistics, including Figure 13 The AUC is described in section B. The optimal folded linear regression models are sorted in descending order of accuracy. Given that the three linear regression models for each set correspond to three randomly selected genes from a list of 62 genes, the most accurate models correspond to the list of genes that underpin those most accurate models. Further analysis (such as applying threshold cutoffs) can further refine the list of informative genes, thereby identifying a number of genes of interest for, for example, biomedical or clinical applications.

[0266] Figure 15 A to Figure 15 B shows the results of deploying the optimally folded linear regression model (as identified based on the method described in this embodiment) on an external dataset. The external dataset is from Kirkeby et al., (2017). Cell Stem Cell 20(1):135-148. The external dataset consisted of cells classified as high or low dopamine levels based on the number of tyrosine hydroxylases per 100,000 transplanted cells (tyrosine hydroxylase is an enzyme essential for dopamine synthesis and is therefore a biomarker for identifying dopaminergic cells). Gene expression data from cells with high dopamine levels were used as input to seven optimal models, each corresponding to the optimal model derived from each of the seven training data portions (i.e., the optimal model derived from each of the seven folds of cross-validation training data). For each of the seven models, the transplantation outcome was predicted by either calculating the average graft size from the three genes used in one of the seven models, or calculating the minimum graft size from the three genes used in one of the seven models. Thus, the AUC for each of the seven models was determined for the case using the average graft size from the three genes and for the case using the minimum graft size from the three genes. When using the average predicted graft size, the average AUC is depicted as... Figure 15 In model A, the average AUC across the seven models is 0.85204. The average AUC when using the smallest predicted graft size is plotted on... Figure 15 In B, the average AUC is 1.00.

[0267] A key advantage of the method described in this embodiment is that it can sort batches based on their predicted ability to survive transplantation in the host brain and establish functional connectivity. Figure 16 A to Figure 16 F illustrates the degree to which the model generated using the method described in this embodiment can distinguish the ordering of graft sizes. For simplicity, results relating to the 3-gene-based model are shown here. Figure 16 A to Figure 16 F shows the normalized graft size for the predicted size and the actual measured size. In six samples (donor 1 to 6) from batches of the same differentiation set that were tested and comparable among them, the magnitude of the difference had a significant impact on the model's ability to resolve the correct rank order. In other words, if the measurement differences between batches are large, the model is able to successfully predict the correct size rank order. The model performs poorly when the measured graft sizes are very similar. This concept is... Figure 16 A to Figure 16 As shown in F. Figure 16 A to Figure 16 The plots in F depict: For each plot, the batch sizes of the two batches from the donor were measured and predicted, and the data were normalized to the maximum value. Circles of fixed radius are overlaid on each plot to provide a visualization of the difference scaling and the model's performance. Gray circles indicate sample differences greater than a radius distance (i.e., a normalized value of 0.2). White circles indicate cases where both samples are within a circle with a radius of 0.2 normalized value. Connecting sample plots ( Figure 16 A to Figure 16 The direction of the line connecting the two data points in each sample plot in F). If the line connecting the data has a positive coefficient, the model is able to correctly distinguish the sorting order (donor 1 ( Figure 16 A), 2 ( Figure 16 B), 3 ( Figure 16 C), 5 ( Figure 16 E), represented by a gray circle. When the line is nearly vertical or has a negative coefficient, the model cannot correctly predict the rank order (donor 4 ( Figure 16 D) and 6 ( Figure 16F), indicated by white circles. These cases relate to the magnitude of the difference between large and small grafts. When the data (on a normalized scale) are very similar in magnitude, the two data points are within a 0.2 radius circle of the largest graft, and these cases cannot be correctly distinguished. When one data point is inside the circle and the other is outside, the model always correctly predicts the rank order. In short, regarding the model's ability to predict the probability of DANPC batches being transplanted into the host, the model performs better when the two DANPC batches are further separated in the space between predicted and measured graft sizes. In other words, in the space between predicted and measured graft sizes, the model generated by the method described in this embodiment is more likely to correctly detect large-value differences and less likely to correctly detect small-value differences.

[0268] Figure 17 A to Figure 17 C shows microscopic images of the graft nuclei of three pairs of DANPC clones 21 days after transplantation. For each pair of clones, plots and analyses were performed corresponding to their predicted and actual graft sizes (e.g., [images would be inserted here]). Figure 16 A pair of clones (as shown). More specifically, Figure 17 The two clones described in A correspond to Figure 16 The two clones described in A for donor 1; Figure 17 The two clones described in B correspond to Figure 16 B describes two clones for donor 2; and Figure 17 The two clones described in C correspond to Figure 16 The two clones described in C for donor 3.

[0269] The method described in this embodiment can be performed with different parameter values ​​and different and / or additional downstream filtering steps. For example, 20,000 random selections of 3 genes from 62 possible genes do not need to be 20,000 random selections. The number of random selections can be changed to, for example, 20,000 random selections of 3 genes from 62 possible genes. In a variation of the method described above, 20,000 random selections of 3 genes are selected from the 62 possible genes identified above. According to the method described above, in each of the 20,000 random selections of 3 genes, 3 linear regression models are identified and folded together. Then, the 20,000 folded linear regression models are filtered so that only models with an accuracy greater than 0.7 are selected. After filtering, 2,766 models remain. All genes appearing in the 2,766 models are combined and sorted according to their accuracy. Then, the top 50 genes associated with the sorted 2,766 models are identified. The expression levels of the top 50 genes were then quantified by normalizing the count per million (CPM) of the genes of interest (which had been normalized to the CPM of reference genes, such as the housekeeping gene GAPDH). Alternatively, the expression levels were quantified by taking the base-2 logarithm of the CPM of the genes of interest (which had been normalized to the base-2 logarithm of the CPM of reference genes, such as the housekeeping gene GAPDH). The results for the top 50 identified genes are listed in Table E1.

[0270] Table E1. Top 50 genes and their conditional ratios associated with graft cell counts ≥1,000 HuNu. .

[0271] Example 3—PCA-based model for predicting dopamine release A machine learning method was developed to predict dopamine release from DANPC cultured with potassium chloride (KCl) based on DANPC gene expression levels. To this end, bulk gene expression levels were determined from a subset of cells in a human DANPC population differentiated from in vitro pluripotent stem cells (iPSCs). The remaining cell subset from the same human DANPC population was cultured in vitro for approximately 60 days. KCl was then added to the DANPCs to stimulate dopamine (and serotonin) release. The dopamine (and serotonin) release from the cultured DANPCs was obtained by liquid chromatography-mass spectrometry (LC-MS). Thus, population-specific bulk gene expression levels and population-specific dopamine release were obtained from a given human DANPC population. For DANPC populations derived from multiple human subjects, population-specific bulk gene expression levels and population-specific dopamine release were also obtained. The PCA-based model described in this embodiment uses DANPC gene expression levels to predict the amount of dopamine released from cultured DANPCs. Figure 6 A schematic workflow for a PCA-based model to predict graft size using bulk gene expression data from DANPC is depicted; however, the same general workflow can be applied to predict the amount of dopamine released from cultured DANPC. It is noteworthy that the method described herein is not limited to predicting the amount of dopamine released from DANPC, but can be generalized to a) predicting the amount of any released biological compound (e.g., neurotransmitters such as dopamine or serotonin), provided that relevant training data is used; and b) other types that are not DANPC.

[0272] Training and testing data used to train and test the machine learning model included bulk RNA sequencing (RNAseq) data of DANPCs. Bulk gene expression levels were then correlated with the amount of dopamine released by the corresponding DANPC. The strength of the correlation between dopamine release and RNAseq expression of the corresponding DANPC was quantified using metrics such as the Pearson correlation coefficient, based on the Weighted Correlation Network Analysis (WGCNA) package in the R programming language.

[0273] A. The data is divided into training and testing parts based on the donor. The Leave-One-Out (LOO) method divides the resulting data on the correlation between bulk RNAseq data and the corresponding dopamine levels released by DANPC into training and test data portions. This process iterates the same number of times as the number of donors represented in the correlation data, for example, 8 iterations. The LOO method is a cross-validation approach. The purpose of cross-validation is to control for the undesirable possibility that the training data portion may not represent the variable of interest used for prediction, and therefore, the model may fail to generalize because it is trained on unusual data. The training data portion can randomly include very unusual data compositions, and training the model based on such unusual compositions may cause the model to fail to generalize across all or most instances of the variable of interest used for prediction. Cross-validation is used to limit this artifact. Cross-validation requires generating many pairs of training and test data portions. In the case of LOO cross-validation, a single data point is reserved as test data, and the remaining data is treated as training data. In this embodiment, the DANPC originates from 8 subjects, so 8 iterations of LOO are used for cross-validation: 1 subject (e.g., subject A) is used for test data, and the remaining 7 subjects (subjects B through H) are used for training data; 1 subject (e.g., subject B) is used for test data, and the remaining 7 subjects (subjects A, C through H) are used for training data; ...; 1 subject (e.g., subject H) is used for test data, and the remaining 7 subjects (subjects A through G) are used for training data. LOO cross-validation is useful when the sample size of independent and separately distributed (IID) random variables is small. This embodiment treats the relevant sample size as the number of subjects, which differs from many other methods that treat the relevant sample size as the number of cultured cells (e.g., DANPC). Although using the number of subjects as the relevant sample size generally results in a smaller sample size than using the number of cells, choosing the number of subjects as the relevant sample size during cross-validation can improve the predictive ability of the machine learning model. However, some methods choose to use the number of cultured cells as the sample size because doing so often enlarges the sample size, and experimenters may consider a larger sample size unconditionally advantageous. However, using the number of cells as the relevant sample size can often lead to poorer predictive power for machine learning models because, in cases where cells originate from a common donor, cells from the common donor are not true IIDs—cells from the common donor may have more in common than cells from another donor. Furthermore, by treating the relevant sample size as the number of subjects, rather than the number of cultured cells (e.g., cell batches), splitting the data into training and testing portions during cross-validation does not result in cells from a given donor being present in both the training and testing portions.This model design choice is particularly important when using RNAseq data from donor cells, because most unique expression features in donor cell RNAseq data indicate traits specific to the donor subject, rather than more general variables of interest common to most donors, such as dopamine release. For the reasons stated above, the method described in this embodiment chooses to use the number of subjects rather than the number of cells or cell batches as the relevant sample size for LOO, even if doing so may result in fewer iterations for LOO. Therefore, the cross-validation method used in this embodiment is limited to cross-validation methods compatible with small sample sizes, hence LOO is used.

[0274] B. PCA was performed on five randomly sampled genes. For each of the eight iterations, a Loo operation was performed on the correlation data (i.e., correlation data regarding the correlation between the expression level of a given gene from the DANPC bulk RNAseq data and the corresponding dopamine release from the DANPC). The correlation data in the training subset was sorted, and the top 2000 correlation values ​​were selected from the training subset of each iteration. For each training subset, 2000 genes corresponding to the 2000 correlation values ​​were selected, and genes common to all eight iterations were identified. Then, all genes common to all eight iterations were filtered for expression levels greater than 0 log counts per million (log CPM). Three hundred and fifty-eight genes were identified that were common to the training subsets of all eight Loo iterations and remained after the expression filtering. From these 358 identified genes, five genes were randomly selected, and principal component analysis (PCA) was performed on these five genes. Five genes were randomly selected from 358 identified genes, and PCA was then performed on these five randomly selected genes. This process was repeated one million times, resulting in one million PCA results (e.g., the percentage of data variance explained by each principal component generated by each PCA). R was determined for each portion of the training data used to generate each of these one million PCA results. 2 The R-value (i.e., the ratio between the RNAseq levels of the five selected genes and the corresponding dopamine release) 2 The result of one million PCA operations is a dataset consisting of one million rows, where each row includes the corresponding R value. 2 The values, and the percentage of data variance explained by each principal component (PC) generated by each PCA. It is worth noting that running one million PCA sessions represents a computational efficiency improvement compared to the full number of PCA sessions required to study each possible selection of five genes from 358 identified genes. The total number of possible combinations of five genes drawn from 358 possible genes is... = 7,648,760,726 possible combinations. Therefore, performing PCA on only one million sets of five genes randomly from these 7,648,760,726 possible combinations saves approximately six orders of magnitude in computational efficiency. Furthermore, selecting rarely used genes (e.g., five genes) from a total of 358 possible genes controls for other significant influences that a particular donor subject will have when training machine learning models (e.g., models derived from PCA). If a large number of genes were selected from a total of 534 possible genes, and PCA were then performed on these large numbers of genes, most of the variance captured by PCA would be unrelated to general dopamine release (i.e., dopamine release common to or unknown to donor subjects) and would instead be primarily related to traits specific to the donor subject. Therefore, to avoid overfitting PCA to the training data, only a small number of genes (e.g., five genes) are repeatedly selected from the 358 possible genes.

[0275] C. A linear regression model was generated based on PC1 and the known dopamine release level. Each row of the 1 million-row dataset describing the PCA results is used to predict dopamine release based on the test portion via linear regression. Some results from each of the 1 million PCA operations can be described as a matrix of 45 cell batches × 5 PCs, where only the column vector describing the results of the first PC (i.e., PC1) (i.e., 45 cell batches × PC1) is of interest. More specifically, the elements of the 45 cell batches × PC1 vector are projections of gene expression (e.g., batch RNAseq) data onto PC1 (i.e., the first set of coordinates in PC space). The 45 projections onto PC1 are plotted against the dopamine releases of the 45 corresponding cell batches from the training portion, and linear regression is performed, i.e., generating the form: The predictive linear model, where y This is the predicted dopamine release. The test data portion is then fed into a predictive linear model (based on the training data portion) to generate the predicted dopamine release. Therefore, each PCA corresponds to a set of five genes from 358 possible genes via a linear relationship, and the resulting predicted dopamine release. Notably, the R-squared value describing the fit of the linear regression to 45 data points is also calculated. 2 The method described in this embodiment generated 1 million PCA operations, which provided a basis for 1 million predictive linear regressions (i.e., regressing the data projected onto PC1 against known dopamine release from the training portion), which predicted 1 million dopamine releases.

[0276] D. Model grouping, folding, and sorting Then, the best model was selected from 1 million linear regression models. For this purpose, x-axis values ​​(PC1%) were generated, including the percentage of variance explained by PC1 and R-axis values. 2 The y-axis grid values. PC1% ranges from 0% to 100%, and R... 2 The values ​​range from 0 to 1. The predicted dopamine release is calculated using a) PC1% values ​​between 0% and 10% and b) R values ​​between 0 and 0.1. 2 Up to 1000 linear regression models were examined to provide the basis for the values, and the predicted dopamine releases were averaged, with cross-validation statistics such as AUC (area under the receiver-operator curve) calculated. In this way, the values ​​were calculated based on their corresponding PC1% values ​​and R... 2 The system groups up to 1000 linear regression models and folds those models together by averaging their predicted dopamine release and calculating individual cross-validation summation statistics for the folded linear regression models, such as individual accuracy and AUC values ​​(and R-squared values). 2 (Sensitivity, specificity, precision, recall, and F1 score). Averaging the predicted dopamine release buffers most of the differences seen in the predicted dopamine release across the various models.

[0277] Repeat the process described above, but use different ranges of PC1% and R. 2 The values ​​were used to group up to 1000 linear regression models. Specifically, the PC1% range increased by 2.5 percentage points. Therefore, the second group of up to 1000 linear regression models included a) PC1% ranging from 2.5% to 12.5%, and b) R values ​​remaining between 0 and 0.1. 2 Value. Similarly, collapse at the specified PC1% and R in the following way. 2 Up to 1000 linear regression models within the range of values: averaging the predicted dopamine release from the models, and calculating individual cross-validation statistics for all folded models, such as individual accuracy and AUC values, as well as the other metrics mentioned above. The groups of up to 1000 linear regression models continue to increase along the x-axis of the grid in increments of 2.5% (and within the range of PC1% and R...). 2At each distinct group within the range, the predicted dopamine release is averaged, and the cross-validation statistic is calculated across the entire group, until the PC1% range becomes at least 90% (assuming a window size of 10%, the window will not move more than 90%, otherwise the effective window size will become less than 10%, and reasonable comparisons with other collapsed models will not be possible). Then, the next group of up to 1000 linear regression models includes a) PC1% ranges between 2.5% and 12.5%, and b) R-values ​​between 0.025 and 0.125. 2 The values ​​were calculated, and similarly, the predicted dopamine releases from up to 1000 models were averaged, and the cross-validation statistic was calculated across the entire group. Based on the PC1% value and R... 2 The iterative process of grouping up to 1000 linear regression models by the range of values ​​continues until the PC1% value is between 0% and 100% and the R-value is between 0 and 1. 2 The entire space of values ​​is filled. As seen above, for each group in the linear regression model, increasing its PC1% value by 2.5 percentage points or its R-value... 2 The value increases by 0.025 (i.e., both the x-axis and y-axis increase by 2.5%). Once the PC1% value and R... 2 If the entire space of values ​​is filled, then up to 1,000 groups of folded linear regression models can be compared via group statistics of interest (e.g., cross-validation statistics).

[0278] It is worth noting that when up to 1000 models are folded, the predicted dopamine release is averaged. Strictly speaking, averaging the predicted dopamine release is not necessary, but it can expose it to different forms of central tendency, such as the median. For any outliers present in the data from the cell batch, the median-based predicted dopamine release may be more robust than the mean-based prediction.

[0279] Figure 18 A to Figure 18 H shows the PC1% and R based on up to 1000 linear regression models. 2 The values ​​are used to group these linear regression models. Figure 18 A to Figure 18 Each plot in H is a heatmap drawn relative to the group statistic of interest, where the x-axis is PC1% and the y-axis is the R-squared value of the linear regression model. 2 ,Right now, Figure 18 A to Figure 18 Each plot in H depicts the PC1%-R described above. 2 Spatial grid. That is, Figure 18 A to Figure 18 H groups the R values ​​of the model in the corresponding order. 2Heatmaps were plotted using AUC, sensitivity, specificity, precision, F1, accuracy, and kappa value. Figure 18 A to Figure 18 In H, the black outline circle in each plot indicates the model with the highest AUC and highest accuracy (from those with similar PC1% and R). 2 (The values ​​are derived from the folded model group). This embodiment sorts the best group of up to 1000 linear regression models, first by the highest accuracy, and then by the highest AUC.

[0280] Figure 19 A to Figure 19 C shows the predictive properties of a set of linear regression models. Figure 19 A describes the relationship between the predicted dopamine release (according to an arbitrary scale normalized to arbitrary units from 0 to 1) and the measured dopamine release (also according to an arbitrary scale normalized to arbitrary units from 0 to 1) for the (folded) model that first has the highest AUC and then the highest accuracy. Figure 19 B describes the receiver-operator characteristic curve (ROC curve) and its corresponding AUC for the model with the highest AUC and the highest accuracy. Figure 19 C depicts a legend including various summary statistics associated with the model that has the highest AUC and the highest accuracy.

[0281] The highest-ranked group of up to 1000 linear regression models can be selected based on any desired criteria, allowing the identification of a desired number of unique genes from the corresponding 5-gene set of those model groups.

[0282] Example 4—A non-PCA-based model for predicting dopamine release A machine learning method was developed to predict dopamine release from DANPC cultured with potassium chloride (KCl) based on DANPC gene expression levels. To this end, bulk gene expression levels were determined from a subset of cells in a human DANPC population differentiated from in vitro pluripotent stem cells (iPSCs). The remaining cell subset from the same human DANPC population was cultured in vitro for approximately 60 days. KCl was then added to the DANPCs to stimulate dopamine (and serotonin) release. The dopamine (and serotonin) release from the cultured DANPCs was obtained by liquid chromatography-mass spectrometry (LC-MS). Thus, population-specific bulk gene expression levels and population-specific dopamine release were obtained from a given human DANPC population. For DANPC populations derived from multiple human subjects, population-specific bulk gene expression levels and population-specific dopamine release were also obtained. The PCA-based model described in this embodiment uses DANPC gene expression levels to predict the amount of dopamine released from cultured DANPCs. Figure 11 A schematic workflow for a non-PCA-based model to predict graft size using bulk gene expression data from DANPCs is depicted; however, the same general workflow can be applied to predict the amount of dopamine released from cultured DANPCs. It is noteworthy that the method described herein is not limited to predicting the amount of dopamine released from DANPCs, but can be generalized to a) predicting the amount of any released biological compound (e.g., neurotransmitters such as dopamine or serotonin), provided that relevant training data is used; and b) other types that are not DANPCs.

[0283] The training and testing data used to train and test the machine learning model included batch RNA sequencing (RNAseq) data of DANPCs. Then, the batch gene expression levels were modeled relative to the dopamine release of the corresponding DANPCs using the edgeR package in the R programming language. The edgeR package models the relationship between individual gene expression and the response variable "dopamine release" using an overly discrete Poisson model, such that each gene receives its own estimate from a generalized linear model likelihood ratio test. Based on the overly discrete Poisson model, for a given training subset, the top 1000 genes associated with dopamine release were identified.

[0284] A. The data is divided into training and testing parts based on the donor. The Leave-One-Out (LOO) method divides the resulting data on the association between bulk RNAseq data and the corresponding DANPC dopamine release into training and testing data portions. This process iterates the same number of times as the number of donors represented in the edgeR modeling data, for example, 8 iterations. The LOO method is a cross-validation approach. The purpose of cross-validation is to control for the undesirable possibility that the training data portion may not represent the variable of interest used for prediction, and therefore, the model may fail to generalize because it is trained on unusual data. The training data portion can randomly include very unusual data compositions, and training the model based on such unusual compositions may cause the model to fail to generalize across all or most instances of the variable of interest used for prediction. Cross-validation is used to limit this artifact. Cross-validation requires generating many pairs of training and testing data portions. In the case of LOO cross-validation, a single data point is reserved as test data, and the remaining data is treated as training data. In this embodiment, DANPC originates from 8 subjects, so 8 iterations of LOO are used for cross-validation: 1 subject (e.g., subject A) for test data, and the remaining 7 subjects (subjects B through H) for training data; 1 subject (e.g., subject B) for test data, and the remaining 7 subjects (subjects A, C through H) for training data; ...; 1 subject (e.g., subject H) for test data, and the remaining 7 subjects (subjects A through G) for training data. LOO cross-validation is useful when the sample size of independent and separately distributed (IID) random variables is small. This embodiment treats the relevant sample size as the number of subjects, which differs from many other methods that treat the relevant sample size as the number of cultured cells or cell batches (e.g., DANPC). Although using the number of subjects as the relevant sample size generally results in a smaller sample size than using the number of cells, choosing the number of subjects as the relevant sample size during cross-validation can improve the predictive ability of the machine learning model. However, some methods choose to use the number of cells as the sample size because doing so often enlarges the sample size, and experimenters may consider a larger sample size unconditionally advantageous. However, using the number of cells or cell batches as the relevant sample size can often lead to poorer predictive power for machine learning models because, in cases where the cultured cells or cell batches originate from a common donor, cells from the common donor are not true IIDs—cells from the common donor may have more in common than cells from another donor. Furthermore, by treating the relevant sample size as the number of subjects rather than the number of cultured cells, splitting the data into training and testing portions during cross-validation does not result in cells from a given donor being present in both the training and testing portions.This model design choice is particularly important when using RNAseq data from donor cells because most unique expression features in donor cell RNAseq data indicate traits specific to the donor subject, rather than more general variables of interest common to most donors, such as neurotransmitter release levels, for example, dopamine release. For the reasons stated above, the method described in this embodiment chooses to use the number of subjects rather than the number of cultured cells or cell batches as the relevant sample size for cross-validation, even if doing so may result in fewer iterations for cross-validation. Therefore, the cross-validation method used in this embodiment is limited to cross-validation methods compatible with small sample sizes, and thus LOO is used.

[0285] B. Calculations were performed on three randomly sampled genes. Based on edgeR-based modeling of gene expression data and their association with dopamine release, the top 1000 genes associated with dopamine release were identified for each of the eight training parts. Then, intersecting genes across each of the eight training parts were identified. In addition to gene intersecting, each gene needed to have at least 40 reads mapped to it in at least six training part RNAseq libraries. Based on these criteria, 173 intersecting genes were identified. These 173 genes were then further filtered using 40 reads mapped to a given gene in at least six training part RNAseq libraries, retaining only the more highly expressed genes. After filtering for highly expressed genes, 140 genes remained. Figure 11 A schematic workflow for a non-PCA-based model to predict dopamine release using bulk gene expression data from DANPC is described.

[0286] Three genes are randomly sampled from the remaining 140 genes. The total number of possible combinations of the three genes drawn from the 140 possible genes is: = 447,580 possible combinations. To achieve greater computational efficiency, only a set of 3 genes was randomly sampled 20,000 times from the 447,580 possibilities, thus improving computational efficiency by nearly two orders of magnitude. Furthermore, selecting rarely used genes (e.g., 3 genes) from the total of 140 possible genes controls for other significant influences that a particular donor subject will have when generating one or more models for predicting dopamine release. If a large number of genes were selected from the total of 140 possible genes, much of the predictive power based on gene expression data would be unrelated to general dopamine release (i.e., dopamine release common to or unknown to donor subjects) and would instead be primarily related to the specific traits of the donor subject. Therefore, to avoid generating models based on overfitting training data, only a small number of genes (e.g., 3 genes) are repeatedly selected from the 140 possible genes.

[0287] C. A linear regression model was generated based on gene expression and known dopamine release levels. To generate a model predicting graft size for each of the three randomly sampled gene combinations, a linear regression was performed between the gene expression level of each of the three genes and the graft size of 45 cell batches, in the form of... The linear regression models were used. To recap, the dataset analyzed in this embodiment comprised 45 cell batches from 7 donor subjects, each with corresponding batch RNAseq data including expression data for the first 140 genes. Therefore, each linear regression model was derived from the sample size of 45 cell batches, where the predictor variable was the batch gene expression of one of three randomly selected genes from a given cell batch, and the response variable was the dopamine release level of a given cell batch.

[0288] Figure 20 A to Figure 20 C describes a linear regression model for each of three randomly selected genes from a list of 140 genes. Figure 20 A to Figure 20 In C, the x-axis of each plot represents the gene expression of one of three randomly selected genes (in this case, GPR35). Figure 20 A), SLC25A37 ( Figure 20 B) and CAMK2N1 Figure 20C), and the y-axis of each plot represents the dopamine release of a cell batch. Each data point on each plot corresponds to one of 45 cell batches. For emphasis, each linear regression model in this embodiment is based on data from the training portion. The linear regression model can then be used to predict dopamine release based on gene expression data from the test portion. For each set of three linear regression models corresponding to three randomly selected sets of genes, the three predicted dopamine releases are averaged to generate a single predicted dopamine release. It is worth noting that the average can be replaced by some other type of central tendency calculation, such as the median, which tends to be more robust than the mean for outliers. The three linear regression models for each set are folded together, and for the folded model, a single accuracy, a single AUC, and other single cross-validation summary statistics are determined. Figure 21 A describes the measured dopamine release and its relationship with the highest R. 2 The relationship between the predicted dopamine release and the model values, and Figure 21 B describes the targets based on GPR35, SLC25A37 and CAMK2N1 (as shown in...). Figure 20 A to Figure 20 The receiver operating characteristic (ROC) curves and their associated AUC values ​​for the 3-gene model (shown in Figure C). Figure 21 C is a legend depicting the multiple summary statistics (including AUC) associated with the folded model. The best folded linear regression models are sorted in descending order of accuracy. Given that the three linear regression models for each set correspond to three randomly selected genes from a list of 62 genes, the most accurate models correspond to the list of genes that underpin those most accurate models. Further analysis (such as applying threshold cutoffs) can further refine the list of informative genes, thereby identifying a number of genes of interest for, for example, biomedical or clinical applications.

[0289] The number of genes of interest can be based on a threshold cutoff value, such as a model accuracy of 0.8. 20,000 models were filtered based on a random selection of 3 genes from 140 possible genes, ensuring that only models with an accuracy greater than 0.8 were selected. After selection, 1406 models remained. All genes appearing in these 1406 models were counted and sorted according to frequency. Then, the top 50 genes associated with the sorted 1406 models, corresponding to dopamine release of at least 15 nM / 10⁻⁶, were identified. 5The expression levels of the top 50 genes were then quantified by normalizing the count per million (CPM) of the genes of interest (which had been normalized to the CPM of reference genes, such as the housekeeping gene GAPDH). Alternatively, the expression levels were quantified by taking the base-2 logarithm of the CPM of the genes of interest (which had been normalized to the base-2 logarithm of the CPM of reference genes, such as the housekeeping gene GAPDH). The results for the top 50 identified genes are listed in Table E2.

[0290] Table E2. DA release ≥15 nM / 10 5 The top 50 genes associated with each cell and their conditional ratio values .

[0291] The scope of this invention is not intended to be limited to the specific embodiments disclosed, which are provided, for example, to illustrate various aspects of the invention. Various modifications to the compositions and methods will become apparent from the description and teachings herein. Such changes may be made without departing from the true scope and spirit of this disclosure and are intended to fall within its scope.

Claims

1. A method for predicting whether a population of neuronal progenitor cells is likely to successfully transplant and survive when implanted into a brain region, the method comprising: (a) Determine the gene expression levels of one or more genes (G genes) associated with predicted transplant viability in a test sample comprising a population of neuronal progenitor cells, wherein the one or more G genes are selected from the group consisting of: AC000120.3, KRT77, TTR, PRR16, MEGF10, PDE3A, GDPD2, CMTM8, APOA1, CMTM7, CDHR3, CORIN, VTN, CPNE8, EFEMP1, CD47, SPARC, JAM2, CDO1, PLXDC2, DYNL L2, ITGA3, RPS6KL1, CHRNB2, SULT4A1, PTPN3, LZTS1, RUNX1T1, TMEM145, EPHA10, CARMIL3, MANEAL, TMEM176B, MPP3, DRAX IN, ADGRB1, KIF26A, CELF5, CNTN2, ASPHD1, SVOP, ANGPT2, SLC22A15, SRRM3, GRIN2D, DACH2, CHST1, GRIN1, LHX5 and NOS2; and (b) Predicting neuronal transplant viability of the neuronal progenitor cells by correlating determined gene expression levels of one or more G genes in the test sample with a reference plot of each G gene, the reference plot correlating graft size with gene expression levels of the G genes in a training set including one or more reference samples.

2. The method of claim 1, wherein each data point on the reference plot is determined by the following steps: (a) Measure the gene expression level of the G gene in a reference sample comprising a population of neuronal progenitor cells; (b) Implanting neuronal progenitor cells from the reference sample into a brain region of a test animal, and measuring the size of the graft formed by the implanted neuronal progenitor cells after the incubation period; and (c) Plot the size of the graft against the expression level of the G gene to obtain data points for the training samples.

3. The method of claim 2, wherein the reference plotting comprises a plurality of data points obtained for each of a plurality of reference samples.

4. The method according to claim 2, wherein the reference plot is obtained by difference expression analysis or linear regression analysis of the plurality of data points.

5. The method of claim 2, wherein the reference plot is obtained by applying the gene expression levels of the one or more G genes in the test sample as input to a machine learning model configured to predict whether a population of neuronal progenitor cells is likely to successfully transplant and survive when implanted into a brain region of a subject, wherein the machine learning model is trained using the gene expression levels of the G genes in multiple reference populations of neuronal progenitor cells.

6. The method of claim 5, wherein the machine learning model includes principal component analysis.

7. The method of claim 2, wherein the predicted graft size is expressed as the number of neurons derived from the neuronal progenitor cells within a cross-section of the brain region of the test animal.

8. The method of claim 7, wherein the brain region is the substantia nigra, and the cross-section of the substantia nigra comprises approximately one-sixth of the substantia nigra.

9. The method of claim 8, wherein if the predicted graft size is greater than 1,000 cells in the cross-section of the substantia nigra, it indicates a high graft survival rate.

10. The method of claim 1, wherein the predicted transplant viability is determined for two or more G genes, and the overall transplant viability prediction for the test sample is based on a combined assessment of the predicted transplant viability obtained for each of the two or more G genes.

11. The method of claim 10, wherein the combined assessment includes determining the mean or median predicted transplant survival rate.

12. The method of claim 1, wherein the one or more G genes are selected from the group consisting of: TTR, PRR16, CMTM8, APOA1, CD47, CD01, KIR26A and CNTN2.

13. The method of claim 12, wherein the one or more G genes are selected from the group consisting of TTR, PRR16 and CD47.

14. The method of claim 13, wherein the one or more G genes are TTR, PRR16, and CD47.

15. The method of claim 1, wherein the population of neuronal progenitor cells is derived from cell cultures differentiated from said cells under conditions that induce neural differentiation of pluripotent stem cells.

16. The method of claim 15, wherein the pluripotent stem cell is an induced pluripotent stem cell (iPSC).

17. The method of claim 15, wherein the pluripotent stem cells are autologous to the subject to be implanted with the neuronal progenitor cells.

18. The method of claim 15, wherein the differentiation conditions include adherent cell culture, and the neuronal progenitor cell test sample is obtained between day 18 and day 24 after the start of the differentiation process.

19. The method of claim 15, wherein the differentiation conditions include suspension cell culture, and the neuronal progenitor cell test sample is obtained between day 13 and day 20 after the start of the differentiation process.

20. The method of claim 19, wherein the neuronal progenitor cell test sample is obtained approximately 16 days after the start of the differentiation process.

21. The method of claim 1, wherein the gene expression level of each of the at least one G gene is determined by RNA sequencing (RNAseq).

22. The method of claim 1, wherein the gene expression level of each of the at least one G gene is determined by polymerase chain reaction (PCR).

23. The method of claim 22, wherein the PCR is quantitative PCR (qPCR).

24. The method of claim 23, wherein the gene expression level of said one or more G genes is determined by the following steps: (a) Obtaining RNA samples from the neuronal progenitor cell test samples; (b) Synthesize complementary DNA from the RNA sample using reverse transcription; (c) Amplifying a specific nucleic acid fragment corresponding to the G gene using quantitative polymerase chain reaction (qPCR), wherein the qPCR comprises using a pair of primers specific to the G gene and optionally a probe specific to the G gene; and (d) Determine the expression level of the G gene based on a normalized quantitative amount.

25. The method of claim 1, wherein the gene expression level of the one or more G genes is normalized to a ratio of the relative expression levels of the G genes and the housekeeping genes, optionally wherein the housekeeping gene is GAPDH.

26. The method of claim 1, wherein the expression levels of at least 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19 or 20 G genes are determined.

27. The method of claim 1, wherein the reference sample comprises a merged sample of neuronal progenitor cells derived from multiple donors.

28. A method for predicting whether a population of neuronal progenitor cells is likely to successfully transplant and survive when implanted into a brain region, the method comprising: (a) Determine the gene expression levels of one or more genes (G genes) associated with transplant survival in a test sample comprising a population of neuronal progenitor cells, wherein the one or more G genes are selected from the group consisting of: AC000120.3, KRT77, TTR, PRR16, MEGF10, PDE3A, GDPD2, CMTM8, APOA1, CMTM7, CDHR3, CORIN, VTN, CPNE8, EFEMP1, CD47, SPARC, JAM2, CDO1, PLXDC2, DYNLL2 ITGA3, RPS6KL1, CHRNB2, SULT4A1, PTPN3, LZTS1, RUNX1T1, TMEM145, EPHA10, CARMIL3, MANEAL, TMEM176B, MPP3, DRAXIN, ADGRB1, KIF26A, CELF5, CNTN2, ASPHD1, SVOP, ANGPT2, SLC22A15, SRRM3, GRIN2D, DACH2, CHST1, GRIN1, LHX5, and NOS2; and (b) The expression level of each of the one or more G genes in the test neuronal progenitor cell population is compared with a predetermined threshold for the specific G gene, wherein the expression level indicates that the neuronal progenitor cells have high transplant viability if any of the following conditions are met: (c) The value is higher than the predetermined threshold of the G gene; or (d) Below the predetermined threshold of the G gene; "Higher than" or "lower than" is defined by the known biological relevance of the G gene in terms of transplant survival ability.

29. The method of claim 28, wherein the predetermined threshold for the specific G gene is based on the expression level of the G gene in a training sample comprising neuronal progenitor cells known to exhibit high transplant viability when implanted into the brain, and the gene expression level of the G gene in a test sample similar to the expression level of the G gene in the training sample predicts that neurons derived from the neuronal progenitor cells in the test sample have high transplant viability potential.

30. The method of claim 28, wherein the predetermined threshold for the specific G gene is based on the expression level of the G gene in a training sample comprising neuronal progenitor cells known to exhibit low transplant viability when implanted into the brain, and the gene expression level of the G gene in a test sample, which is similar to the expression level of the G gene in the control sample, predicts that neurons derived from the neuronal progenitor cells in the test sample have low transplant viability potential.

31. The method of claim 28, wherein the neuronal progenitor cells are predicted to have high transplant viability after implantation into the brain if the following conditions are met: (a) The gene expression level of at least one first G gene selected from the group consisting of the following is below the predetermined threshold of the first G gene: AC000120.3, KRT77, TTR, PRR16, MEGF10, PDE3A, GDPD2, CMTM8, APOA1, CMTM7, CDHR3, CORIN, VTN, CPNE8, EFEMP1, CD47, SPARC, JAM2, CDO1, and PLXDC2; and / or (b) The expression level of at least one second G gene selected from the group consisting of the following is higher than the predetermined threshold of the second G gene: DYNLL2, ITGA3, RPS6KL1, CHRNB2, SULT4A1, PTPN3, LZTS1, RUNX1T1, TMEM145, EPHA10, CARMIL3, MANEAL, TMEM176B, MPP3, DRAXIN, ADGRB1, KIF26A, CELF5, CNTN2, ASPHD1, SVOP, ANGPT2, SLC22A15, SRRM3, GRIN2D, DACH2, CHST1, GRIN1, LHX5, and NOS2.

32. The method of claim 31, wherein the predetermined threshold for the specific G gene is based on the ratio of the relative expression levels of a) the G gene and b) the control gene in the test sample.

33. The method of claim 32, wherein the control gene is GAPDH, and the predetermined threshold is selected from the group consisting of: (a) a ratio of AC000120.3 to GAPDH expression less than about 0.14; (b) a ratio of KRT77 to GAPDH expression less than about 0.68; (c) a ratio of TTR to GAPDH expression less than about 1.11; (d) a ratio of PRR16 to GAPDH expression less than about 0.43; (e) a ratio of MEGF10 to GAPDH expression less than about 0.79; (f) a ratio of PDE3A to GAPDH expression less than about 1.00; (g) a ratio of GDPD2 to GAPDH expression less than about 0.78; (h) a ratio of CMTM8 to GAPDH expression less than about 1.02; (i) (j) The ratio of APOA1 to GAPDH expression less than about 0.68; (k) The ratio of CMTM7 to GAPDH expression less than about 0.88; (l) The ratio of CDHR3 to GAPDH expression less than about 1.09; (m) The ratio of CORIN to GAPDH expression less than about 1.24; (n) The ratio of VTN to GAPDH expression less than about 0.98; (d) The ratio of CPNE8 to GAPDH expression less than about 0.79; (e) The ratio of EFEMP1 to GAPDH expression less than about 0.83; (p) The ratio of CD47 to GAPDH expression less than about 1.16; (q) The ratio of SPARC to GAPDH expression less than about 1.29; (r) The ratio of JAM2 to GAPDH expression less than about 0.82; (s) The ratio of CDO1 to GAPDH expression less than about 1.00; (t) (u) The ratio of PLXDC2 to GAPDH expression less than about 1.00; (v) The ratio of DYNLL2 to GAPDH expression greater than about 0.56; (v) The ratio of ITGA3 to GAPDH expression greater than about 0.26; (w) The ratio of RPS6KL1 to GAPDH expression greater than about 0.21; (x) The ratio of CHRNB2 to GAPDH expression greater than about 0.23; (y) The ratio of SULT4A1 to GAPDH expression greater than about 0.22; (z) The ratio of PTPN3 to GAPDH expression greater than about 0.03; (aa) The ratio of LZTS1 to GAPDH expression greater than about 0.19; (ab) The ratio of RUNX1T1 to GAPDH expression greater than about 0.24; (ac) The ratio of TMEM145 to GAPDH expression greater than about 0.05; (ad) The ratio of EPHA10 to GAPDH expression greater than approximately 0.16; (ae) the ratio of CARMIL3 to GAPDH expression greater than approximately 0.16; (af) the ratio of MANEAL to GAPDH expression greater than approximately 0.16.The ratios of TMEM176B to GAPDH expression were greater than approximately 0.11; (ah) the ratio of MPP3 to GAPDH expression was greater than approximately 0.12; (ai) the ratio of DRAXIN to GAPDH expression was greater than approximately 0.27; (aj) the ratio of ADGRB1 to GAPDH expression was greater than approximately 0.07; (ak) the ratio of KIF26A to GAPDH expression was greater than approximately 0.23; (al) the ratio of CELF5 to GAPDH expression was greater than approximately 0.25; (am) the ratio of CNTN2 to GAPDH expression was greater than approximately 0.23; (an) the ratio of ASPHD1 to GAPDH expression was greater than approximately 0.08; (ao) the ratio of SVOP to GAPDH expression was greater than approximately 0.16; (ap) the ratio of ANGPT2 to GAPDH expression was greater than approximately 0.06; (aq) The ratios of SLC22A15 to GAPDH expression greater than approximately 0.04; (ar) the ratio of SRRM3 to GAPDH expression greater than approximately 0.17; (as) the ratio of GRIN2D to GAPDH expression greater than approximately 0.02; (at) the ratio of DACH2 to GAPDH expression greater than approximately 0.06; (au) the ratio of CHST1 to GAPDH expression greater than approximately 0.04; (av) the ratio of GRIN1 to GAPDH expression greater than approximately 0.26; (aw) the ratio of LHX5 to GAPDH expression greater than approximately 0.06; and (ax) the ratio of NOS2 to GAPDH expression greater than approximately 0.

08.

34. A method for training a machine learning model, said machine learning model being used to predict whether a population of neuronal progenitor cells is likely to successfully survive when implanted into a brain region, said method comprising: (a) Obtain the gene expression levels of one or more genes in each of multiple reference populations of neuronal progenitor cells; (b) Receive transplant survival adaptation information for each of the plurality of reference populations, wherein the transplant survival adaptation information of the reference population indicates whether the neuronal progenitor cells have transplanted into the brain region of the subject after implantation of the neuronal progenitor cells of the reference population into the brain region or the extent to which the neuronal progenitor cells have transplanted into the brain region. as well as (c) Using the gene expression levels of (a) and the transplant survival adaptation information of (b) as inputs to train a machine learning model, wherein the machine learning model is trained to predict, based on the gene expression levels of the plurality of genes, whether the neuronal progenitor cell population will transplant and survive in the brain region after implantation of the neuronal progenitor cell population into the brain region of the subject.

35. A kit for predicting the likelihood of successful transplantation and survival of a population of neuronal progenitor cells when implanted into a brain region, said kit comprising one or more of the following: (a) The first pair of oligonucleotide primers suitable for amplifying the first gene; (b) A second pair of oligonucleotide primers suitable for amplifying the second gene; and (c) A third pair of oligonucleotide primers suitable for amplifying the third gene; Each of the first, second, and third genes is selected from the group consisting of: AC000120.3, KRT77, TTR, PRR16, MEGF10, PDE3A, GDPD2, CMTM8, APOA1, CMTM7, CDHR3, CORIN, VTN, CPNE8, EFEMP1, CD47, SPARC, JAM2, CDO1, PLXDC2, DYNLL2, ITGA3, RPS6KL1. CHRNB2, SULT4A1, PTPN3, LZTS1, RUNX1T1, TMEM145, EPHA10, CARMIL3, MANEAL, TMEM176B, MPP3, DRAXIN, ADGR B1, KIF26A, CELF5, CNTN2, ASPHD1, SVOP, ANGPT2, SLC22A15, SRRM3, GRIN2D, DACH2, CHST1, GRIN1, LHX5 and NOS2.

36. The kit according to claim 35, wherein the first gene, the second gene and the third gene are each selected from the group consisting of: TTR, PRR16, CMTM8, APOA1, CD47, CD01, KIR26A and CNTN2.

37. The kit of claim 36, wherein the kit comprises at least a first pair of oligonucleotide primers, a second pair of oligonucleotide primers and a third pair of oligonucleotide primers, and the first gene is TTR, the second gene is PRR16 and the third gene is CD47.

38. A computing device configured to predict the transplant survival potential of neuronal progenitor cells when a population of neuronal progenitor cells is implanted into a brain region, the computing device comprising: (a) Processor; (b) A memory including instructions executable by the processor, the instructions being configured to perform the following steps: (i) Receiving test samples comprising gene expression levels of one or more genes (G genes) associated with predicted transplant viability in a population of neuronal progenitor cells, wherein the one or more G genes are selected from the group consisting of: AC000120.3, KRT77, TTR, PRR16, MEGF10, PDE3A, GDPD2, CMTM8, APOA1, CMTM7, CDHR3, CORIN, VTN, CPNE8, EFEMP1, CD47, SPARC, JAM2, CDO1, PLXDC2 , DYNLL2, ITGA3, RPS6KL1, CHRNB2, SULT4A1, PTPN3, LZTS1, RUNX1T1, TMEM145, EPHA10, CARMIL3, MANEAL, TMEM176B, MPP3, DRAXIN, ADGRB1, KIF26A, CELF5, CNTN2, ASPHD1, SVOP, ANGPT2, SLC22A15, SRRM3, GRIN2D, DACH2, CHST1, GRIN1, LHX5, and NOS2; (ii) Determine the gene expression level of each of the one or more G genes based on the test sample; (iii) The determined gene expression level of each of the one or more G genes in the test sample is compared with a reference plot of each corresponding G gene, wherein each reference plot correlates the gene expression level of the G gene with graft size data obtained from a training set including one or more reference samples. as well as (iv) By correlating the determined gene expression levels of one or more G genes in the test sample with the reference plotting data to predict the neuronal transplantation survival ability of the neuronal progenitor cells in the test sample, a predictive assessment of the transplantation survival potential of the neuronal progenitor cell population is generated.

39. A method for predicting whether neurons derived from a population of neuronal progenitor cells will produce dopamine, the method comprising: (a) Determine the gene expression levels of one or more genes (D genes) associated with predicted dopamine production in a test sample comprising a population of neuronal progenitor cells, wherein the D genes are selected from the group consisting of: CNTNAP5, KLHL1, NHLH2, GREM2, BRINP2, GRIN3A, LRRC4C, IRX3, CPNE4, PTPN3, PMEL, PCDH20, LRRC37A2, TMEM246, B3GALNT1, ZHX1, BCAS4, SLC25A37, GRINA, MID1, FRMD4A, PARP1 0. WHAMMP2, EYA1, CORO2B, WHAMMP3, B3GALT5, GPR35, ABCD2, ITIH3, AC107464.1, CAMK2N1, CAMK2A, PRPS1, GOLGA6L10, AMOT, SULT 1A1, CD83, SPON1, FRMPD3, AC096570.1, TCAF2, GOLGA8M, VWA5B2, CA8, AC017050.1, KRT77, AP000350.6, LINC02751 and ARHGAP5-AS1; as well as (b) Predicting the dopamine-producing capacity of neurons derived from the neuronal progenitor cells by associating determined gene expression levels of the one or more D genes in the test samples with a reference map of each D gene, the reference map associating dopamine production of the neurons with gene expression levels of the D genes in a training set including one or more reference samples.

40. The method of claim 39, wherein each data point on the reference plot is determined by the following steps: (a) Measure the gene expression level of the D gene in a reference sample comprising a population of neuronal progenitor cells; (b) Differentiating neuronal progenitor cells from the reference sample into neurons and measuring the amount of dopamine produced by the neurons; and (c) Plot the dopamine production against the expression level of the D gene to obtain data points for the training samples.

41. The method of claim 39, wherein the reference plotting comprises a plurality of data points obtained for each of a plurality of reference samples.

42. The method of claim 39, wherein the reference plot is obtained by difference expression analysis or linear regression analysis of the plurality of data points.

43. The method of claim 39, wherein the reference plot is obtained by applying the gene expression levels of the one or more D genes in the test sample as input to a machine learning model configured to predict whether neurons derived from a population of neuronal progenitor cells will produce dopamine, wherein the machine learning model is trained using the gene expression levels of the D genes in multiple reference populations of neuronal progenitor cells.

44. The method of claim 43, wherein the machine learning model comprises principal component analysis.

45. The method of claim 39, wherein if the predicted amount of dopamine produced by the neuron derived from the neuronal progenitor is at least 15 nM dopamine / 10 5 If a population of neurons is identified, then the neuronal progenitor cell population is predicted to have a high dopamine production capacity.

46. ​​The method of claim 39, wherein the predicted dopamine production capacity is determined for two or more D genes, and the overall dopamine production capacity prediction for the test sample is based on a combined assessment of the predicted dopamine production capacity for each of the two or more D genes.

47. The method of claim 46, wherein the combined assessment includes determining the mean or median predicted transplant viability.

48. The method of claim 39, wherein the one or more D genes are selected from the group consisting of: CNTNAP5, NHLH2, GREM2, PMEL, PCDH20, LRRC37A2, SLC25A37, MID1, EYA1, B3GALT5, GPR35, AC107464.1, CAMK2N1, CAMK2A, GOLGA6L10, FRMPD3, VWA5B2, AC017050.1, and LINC02751.

49. The method of claim 48, wherein the one or more D genes are selected from the group consisting of B3GALT5, FRMPD3 and GREM2.

50. The method of claim 49, wherein the one or more D genes are B3GALT5, FRMPD3, and GREM2.

51. The method of claim 39, wherein the neuronal progenitor cell population is derived from a cell culture differentiated from said cells under conditions that induce neural differentiation of pluripotent stem cells.

52. The method of claim 51, wherein the pluripotent stem cell is an induced pluripotent stem cell (iPSC).

53. The method of claim 51, wherein the pluripotent stem cells are autologous to the subject to be implanted with the neuronal progenitor cells.

54. The method of claim 51, wherein the differentiation conditions include adherent cell culture, and the neuronal progenitor cell test sample is obtained between day 18 and day 24 after the start of the differentiation process.

55. The method of claim 51, wherein the differentiation conditions include suspension cell culture, and the neuronal progenitor cell test sample is obtained between day 13 and day 20 after the start of the differentiation process.

56. The method of claim 55, wherein the neuronal progenitor cell test sample is obtained approximately 16 days after the start of the differentiation process.

57. The method of claim 39, wherein the gene expression level of each of the at least one D gene is determined by RNA sequencing (RNAseq).

58. The method of claim 39, wherein the gene expression level of each of the at least one D gene is determined by polymerase chain reaction (PCR).

59. The method of claim 58, wherein the PCR is quantitative PCR (qPCR).

60. The method of claim 59, wherein the gene expression level of the at least one D gene is determined by the following steps: (a) Obtaining RNA samples from the neuronal progenitor cell test samples; (b) Synthesize complementary DNA from the RNA sample using reverse transcription; (c) Amplifying a specific nucleic acid fragment corresponding to the D gene using quantitative polymerase chain reaction (qPCR), wherein the qPCR comprises using a pair of primers specific to the D gene and optionally a probe specific to the D gene; and (d) Determine the expression level of the D gene based on a normalized quantitative amount.

61. The method of claim 39, wherein the gene expression level of the one or more D genes is normalized to a ratio of the relative expression levels of the D genes and the housekeeping genes, optionally wherein the housekeeping gene is GAPDH.

62. The method of claim 39, wherein the expression levels of at least 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19 or 20 D genes are determined.

63. The method of claim 39, wherein the reference sample comprises a merged sample of neuronal progenitor cells derived from multiple donors.

64. A method for predicting whether neurons derived from a population of neuronal progenitor cells will produce dopamine, the method comprising: (a) Determine the gene expression levels of one or more genes (D genes) associated with predicted dopamine production capacity in a test sample comprising a population of neuronal progenitor cells, wherein the one or more D genes are selected from the group consisting of: CNTNAP5, KLHL1, NHLH2, GREM2, BRINP2, GRIN3A, LRRC4C, IRX3, CPNE4, PTPN3, PMEL, PCDH20, LRRC37A2, TMEM246, B3GALNT1, ZHX1, BCAS4, SLC25A37, GRINA, MID1, FRMD4A, PA RP10, WHAMMP2, EYA1, CORO2B, WHAMMP3, B3GALT5, GPR35, ABCD2, ITIH3, AC107464.1, CAMK2N1, CAMK2A, PRPS1, GOLGA6L10, AMOT, SUL T1A1, CD83, SPON1, FRMPD3, AC096570.1, TCAF2, GOLGA8M, VWA5B2, CA8, AC017050.1, KRT77, AP000350.6, LINC02751 and ARHGAP5-AS1; as well as (b) The expression level of each of the one or more D genes in the test neuronal progenitor cell population is compared with a predetermined threshold for the specific D gene, wherein the expression level indicates that the neuronal progenitor cells have a high predicted dopamine production capacity if any of the following conditions are met: (c) Above the predetermined threshold of the D gene; or (d) Below the predetermined threshold of the D gene; "Higher than" or "lower than" is defined by the known biological relevance of the G gene in the context of its predicted dopamine production capacity.

65. The method of claim 64, wherein the predetermined threshold for the specific D gene is based on the expression level of the D gene in a training sample comprising neuronal progenitor cells known to generate neurons producing high levels of dopamine, and the gene expression level of the D gene in a test sample similar to the expression level of the D gene in the training sample predicts that neurons derived from the neuronal progenitor cells in the test sample have high dopamine-producing potential.

66. The method of claim 64, wherein the predetermined threshold for the specific D gene is based on the expression level of the D gene in a training sample comprising neuronal progenitor cells known to generate neurons producing low levels of dopamine, and the gene expression level of the D gene in a test sample, similar to the expression level of the D gene in the control sample, predicts that neurons derived from the neuronal progenitor cells in the test sample have low dopamine-producing potential.

67. The method of claim 64, wherein the neuronal progenitor is predicted to produce neurons with high dopamine production capacity if the following conditions are met: (a) The gene expression level of at least one first D gene selected from the group consisting of the following is below the predetermined threshold of the first D gene: CNTNAP5, KLHL1, NHLH2, GREM2, BRINP2, GRIN3A, LRRC4C, IRX3, CPNE4, PTPN3, PMEL, PCDH20, LRRC37A2, TMEM246, B3GALNT1, and ZHX1; and / or (b) The gene expression level of at least one second D gene selected from the group consisting of the following is higher than the predetermined threshold of the second D gene: BCAS4, SLC25A37, GRINA, MID1, FRMD4A, PARP10, WHAMMP2, EYA1, CORO2B, WHAMMP3, B3GALT5, GPR35, ABCD2, ITIH3, AC107464.1, CAMK2N1, CAMK2A, PRPS1, GOLGA6L10, AMOT, SULT1A1, CD83, SPON1, FRMPD3, AC096570.1, TCAF2, GOLGA8M, VWA5B2, CA8, AC017050.1, KRT77, AP000350.6, LINC02751, and ARHGAP5-AS1.

68. The method of claim 67, wherein the predetermined threshold for the specific D gene is based on the ratio of the relative expression levels of a) the D gene and b) the control gene in the test sample.

69. The method of claim 68, wherein the control gene is GAPDH, and the predetermined threshold is selected from the group consisting of: (a) a ratio of CNTNAP5 to GAPDH expression less than about 0.12; (b) a ratio of KLHL1 to GAPDH expression less than about 0.10; (c) a ratio of NHLH2 to GAPDH expression less than about 0.56; (d) a ratio of GREM2 to GAPDH expression less than about 0.35; (e) a ratio of BRINP2 to GAPDH expression less than about 0.97; (f) a ratio of GRIN3A to GAPDH expression less than about 0.48; (g) a ratio of LRRC4C to GAPDH expression less than about 0.39; (h) a ratio of IRX3 to GAPDH expression less than about 0.55; (i) a ratio of CPNE4 to GAPDH expression less than about 0.28; (j) (k) The ratio of PTPN3 to GAPDH expression less than about 0.25; (l) The ratio of PMEL to GAPDH expression less than about 0.29; (m) The ratio of PCDH20 to GAPDH expression less than about 0.20; (n) The ratio of LRRC37A2 to GAPDH expression less than about 0.68; (n) The ratio of TMEM246 to GAPDH expression less than about 0.53; (o) The ratio of B3GALNT1 to GAPDH expression less than about 0.67; (p) The ratio of ZHX1 to GAPDH expression less than about 0.55; (q) The ratio of BCAS4 to GAPDH expression greater than about 0.42; (r) The ratio of SLC25A37 to GAPDH expression greater than about 0.38; (s) The ratio of GRINA to GAPDH expression greater than about 0.60; (t) The ratio of MID1 to GAPDH expression greater than about 0.62; (u) The ratio of FRMD4A to GAPDH expression greater than approximately 0.57; (v) the ratio of PARP10 to GAPDH expression greater than approximately 0.25; (w) the ratio of WHAMMP2 to GAPDH expression greater than approximately 0.37; (x) the ratio of EYA1 to GAPDH expression greater than approximately 0.32; (y) the ratio of CORO2B to GAPDH expression greater than approximately 0.40; (z) the ratio of WHAMMP3 to GAPDH expression greater than approximately 0.34; (aa) the ratio of B3GALT5 to GAPDH expression greater than approximately 0.40; (ab) the ratio of GPR35 to GAPDH expression greater than approximately 0.19; (ac) the ratio of ABCD2 to GAPDH expression greater than approximately 0.35; (ad) the ratio of ITIH3 to GAPDH expression greater than approximately 0.17; (ae) the ratio of AC107464.1 to GAPDH expression greater than approximately 0.The ratio of 20; (af) CAMK2N1 to GAPDH expression greater than approximately 0.52; (ag) CAMK2A to GAPDH expression greater than approximately 0.37; (ah) PRPS1 to GAPDH expression greater than approximately 0.52; (ai) GOLGA6L10 to GAPDH expression greater than approximately 0.21; (aj) AMOT to GAPDH expression greater than approximately 0.50; (ak) SULT1A1 to GAPDH expression greater than approximately 0.18; (al) CD83 to GAPDH expression greater than approximately 0.29; (am) SPON1 to GAPDH expression greater than approximately 0.76; (an) FRMPD3 to GAPDH expression greater than approximately 0.31; (ao) AC096570.1 to GAPDH expression greater than approximately 0.14; (ap) The ratios of TCAF2 to GAPDH expression were greater than approximately 0.30; (aq) the ratio of GOLGA8M to GAPDH expression was greater than approximately 0.003; (ar) the ratio of VWA5B2 to GAPDH expression was greater than approximately 0.22; (as) the ratio of CA8 to GAPDH expression was greater than approximately 0.19; (at) the ratio of AC017050.1 to GAPDH expression was greater than approximately 0.08; (au) the ratio of KRT77 to GAPDH expression was greater than approximately 0.14; (av) the ratio of AP000350.6 to GAPDH expression was greater than approximately 0.31; (aw) the ratio of LINC02751 to GAPDH expression was greater than approximately 0.19; and (ax) the ratio of ARHGAP5-AS1 to GAPDH expression was greater than approximately 0.

26.

70. A method for training a machine learning model, said machine learning model for predicting whether neurons derived from a population of neuronal progenitor cells will produce dopamine, said method comprising: (a) Obtain the gene expression levels of one or more genes in each of multiple reference populations of neuronal progenitor cells; (b) Receive dopamine production information from neuronal cells derived from each of the plurality of reference populations, wherein the dopamine production information of the reference populations indicates whether or to what extent cells derived from the neuronal progenitor cells produce dopamine. as well as (c) Using the gene expression levels of (a) and the dopamine production information of (b) as inputs to train a machine learning model, wherein the machine learning model is trained to predict whether neurons derived from a population of neuronal progenitor cells will produce dopamine based on the gene expression levels of the plurality of genes.

71. A kit for predicting dopamine production in neurons derived from a population of neuronal progenitor cells, said kit comprising one or more of the following: (a) The first pair of oligonucleotide primers suitable for amplifying the first gene; (b) A second pair of oligonucleotide primers suitable for amplifying the second gene; and (c) A third pair of oligonucleotide primers suitable for amplifying the third gene; Each of the first gene, the second gene, and the third gene is selected from the group consisting of: CNTNAP5, KLHL1, NHLH2, GREM2, BRINP2, GRIN3A, LRRC4C, IRX3, CPNE4, PTPN3, PMEL, PCDH20, LRRC37A2, TMEM246, B3GALNT1, ZHX1, BCAS4, SLC25A37, GRINA, MID1, FRMD4A, PARP10, WHAMMP2, EYA1, C ORO2B, WHAMMP3, B3GALT5, GPR35, ABCD2, ITIH3, AC107464.1, CAMK2N1, CAMK2A, PRPS1, GOLGA6L10, AMOT, SULT1A1, CD83 , SPON1, FRMPD3, AC096570.1, TCAF2, GOLGA8M, VWA5B2, CA8, AC017050.1, KRT77, AP000350.6, LINC02751 and ARHGAP5-AS1.

72. The kit according to claim 71, wherein the first gene, the second gene, and the third gene are each selected from the group consisting of: CNTNAP5, NHLH2, GREM2, PMEL, PCDH20, LRRC37A2, SLC25A37, MID1, EYA1, B3GALT5, GPR35, AC107464.1, CAMK2N1, CAMK2A, GOLGA6L10, FRMPD3, VWA5B2, AC017050.1, and LINC02751.

73. The kit according to claim 72, wherein the first gene is B3GALT5, the second gene is GREM2, and the third gene is FRMPD3.

74. A computing device configured to predict whether neurons differentiating from a population of neuronal progenitor cells will produce dopamine, the computing device comprising: (a) Processor; (b) A memory including instructions executable by the processor, the instructions being configured to perform the following steps: (i) Receive test samples comprising gene expression data of one or more genes (D genes) associated with predicted dopamine production potential in a population of neuronal progenitor cells, wherein the D genes are selected from the group consisting of: CNTNAP5, KLHL1, NHLH2, GREM2, BRINP2, GRIN3A, LRRC4C, IRX3, CPNE4, PTPN3, PMEL, PCDH20, LRRC37A2, TMEM246, B3GALNT1, ZHX1, BCAS4, SLC25A37, GRINA, MID1, FRMD4A, P ARP10, WHAMMP2, EYA1, CORO2B, WHAMMP3, B3GALT5, GPR35, ABCD2, ITIH3, AC107464.1, CAMK2N1, CAMK2A, PRPS1, GOLGA6L10, AMOT, SU LT1A1, CD83, SPON1, FRMPD3, AC096570.1, TCAF2, GOLGA8M, VWA5B2, CA8, AC017050.1, KRT77, AP000350.6, LINC02751 and ARHGAP5-AS1; (ii) Determine the gene expression level of each of the one or more D genes based on the test sample; (iii) The determined gene expression level of each of the one or more D genes in the test sample is compared with a reference plot of each corresponding D gene, wherein each reference plot correlates the gene expression level of the D gene with the dopamine production level obtained from a training set including one or more reference samples. as well as (iv) By correlating the determined gene expression levels of the one or more D genes with the reference plotting data to predict the dopamine-producing capacity of neurons derived from the neuronal progenitor cells in the test sample, a predictive assessment of the dopamine-producing potential of the derived neurons is generated.

75. A potency assay matrix for determining the efficacy of a neuronal progenitor cell population in treating neurodegenerative diseases, the method comprising subjecting the neuronal progenitor cell population to at least two of the following steps (a), (b), and (c): (a) Classifying the in vitro population of neuronal progenitor cells to determine whether the neuronal progenitor cells include specific dopaminergic precursor cells through the following steps: (i) Receive a test dataset as input, the test dataset including the expression levels of one or more genes expressed in a first test sample including the neuronal progenitor cells; (ii) Using the test dataset and the first reference dataset, calculate a first similarity score for the first test sample, wherein: (1) The first reference dataset includes a representation of the gene expression levels of one or more genes that are differentially expressed between cells in a first differentiation state and cells in a second differentiation state, wherein the second differentiation state is a defined differentiation state of dopaminergic neurons, and wherein the first differentiation state is earlier or later than the second differentiation state in the stem cell differentiation pathway. (2) The expression levels in the test dataset include expressions representing the expression levels of one or more genes included in the first reference dataset, and (3) The first similarity score indicates whether the differentiation state of the test cell is more similar to the first differentiation state or the second differentiation state; (iii) Determine a novelty score for the neuronal progenitor cells in the first test sample, wherein the novelty score indicates the degree to which the gene expression levels in the test dataset deviate from the gene expression levels in the reference database; and (iv) Determine whether the first test sample includes the identified dopaminergic neuron cells based on the similarity score and the novelty score; (b) Predict whether the neuronal progenitor cells are likely to successfully survive when implanted into the brain region using the following steps: (i) Determine the gene expression levels of one or more genes (G genes) associated with predicted transplant viability in a second test sample including the said neuronal progenitor cells, wherein the one or more G genes are selected from the group consisting of: AC000120.3, KRT77, TTR, PRR16, MEGF10, PDE3A, GDPD2, CMTM8, APOA1, CMTM7, CDHR3, CORIN, VTN, CPNE8, EFEMP1, CD47, SPARC, JAM2, CDO1, PLXDC2, DYN LL2, ITGA3, RPS6KL1, CHRNB2, SULT4A1, PTPN3, LZTS1, RUNX1T1, TMEM145, EPHA10, CARMIL3, MANEAL, TMEM176B, MPP3, DRAXIN, ADGRB1, KIF26A, CELF5, CNTN2, ASPHD1, SVOP, ANGPT2, SLC22A15, SRRM3, GRIN2D, DACH2, CHST1, GRIN1, LHX5, and NOS2; and (ii) Predicting neuronal transplant viability of the neuronal progenitor cells by correlating determined gene expression levels of one or more G genes in the second test sample with a reference map of each G gene, the reference map correlating graft size with gene expression levels of the G genes in a training set including one or more reference samples; and (c) Predict whether neurons derived from the neuronal progenitor cell population will produce dopamine using the following steps: (i) Determine the gene expression levels of one or more genes (D genes) associated with predicted dopamine production in a third test sample comprising a population of neuronal progenitor cells, wherein the D genes are selected from the group consisting of: CNTNAP5, KLHL1, NHLH2, GREM2, BRINP2, GRIN3A, LRRC4C, IRX3, CPNE4, PTPN3, PMEL, PCDH20, LRRC37A2, TMEM246, B3GALNT1, ZHX1, BCAS4, SLC25A37, GRINA, MID1, FRMD4A, PARP1 0. WHAMMP2, EYA1, CORO2B, WHAMMP3, B3GALT5, GPR35, ABCD2, ITIH3, AC107464.1, CAMK2N1, CAMK2A, PRPS1, GOLGA6L10, AMOT, SULT1 A1, CD83, SPON1, FRMPD3, AC096570.1, TCAF2, GOLGA8M, VWA5B2, CA8, AC017050.1, KRT77, AP000350.6, LINC02751 and ARHGAP5-AS1; and (ii) Predicting the dopamine-producing capacity of neurons derived from the neuronal progenitor cells by associating the determined gene expression levels of the one or more D genes in the third test sample with a reference map of each D gene, the reference map associating the dopamine production of the neurons with the gene expression levels of the D genes in a training set including one or more reference samples.

76. The method of claim 75, wherein the method comprises steps (a) and (b).

77. The method of claim 75, wherein the method comprises steps (a) and (c).

78. The method of claim 75, wherein the method comprises steps (b) and (c).

79. The method of claim 75, wherein the method comprises all three steps in steps (a), (b) and (c).

80. The method of claim 75, wherein the method comprises step (b), and the G gene is selected from the group consisting of: TTR, PRR16, CMTM8, APOA1, CD47, CD01, KIR26A, and CNTN2.

81. The method of claim 80, wherein the one or more G genes are TTR, PRR16, and CD47.

82. The method of claim 75, wherein the method comprises step (c), and the one or more D genes are selected from the group consisting of: CNTNAP5, NHLH2, GREM2, PMEL, PCDH20, LRRC37A2, SLC25A37, MID1, EYA1, B3GALT5, GPR35, AC107464.1, CAMK2N1, CAMK2A, GOLGA6L10, FRMPD3, VWA5B2, AC017050.1, and LINC02751.

83. The method of claim 82, wherein the one or more D genes are B3GALT5, FRMPD3, and GREM2.

84. A therapeutic composition comprising a population of neuronal progenitor cells selected by the method of claim 75.

85. A therapeutic composition comprising neuronal progenitor cells derived from pluripotent stem cells, wherein the therapeutic composition comprises neuronal progenitor cells from at least two populations selected from the group consisting of: (a) A first neuronal progenitor cell population, which is classified as a defined dopaminergic precursor cell population using a method that includes classifying the neuronal progenitor cells based on probability scores and deviation scores. (b) A second neuronal progenitor cell population, predicted to generate neurons with high transplant viability; and (c) A third neuronal progenitor cell population, which is predicted to produce neurons with high dopamine production.

86. The therapeutic composition of claim 85, wherein each of the first neuronal progenitor cell population, the second neuronal progenitor cell population, and the third neuronal progenitor cell population is obtained from the same patient.

87. A treatment method comprising implanting a therapeutically effective amount of the therapeutic composition according to claim 85 into a brain region of a subject suffering from a neurodegenerative disease or condition, optionally wherein the neurodegenerative disease or condition includes loss of dopaminergic neurons.

88. A method for transplanting and surviving neuronal cells in a brain region of a subject, the method comprising implanting a therapeutically effective amount of the therapeutic composition according to claim 85 into a brain region of a subject suffering from a neurodegenerative disease or symptom.

89. A method for increasing dopamine production in a brain region of a subject, the method comprising implanting a therapeutically effective amount of the therapeutic composition according to any one of claims 85 into a brain region of a subject suffering from a neurodegenerative disease or symptom.

Images