Chromatin structure biomarkers

Abstract

A method for determining the chromatin state of a sample from transcript abundance data from the sample is described herein. The method includes the steps of: receiving transcript abundance data for multiple transcripts from the sample; determining a correlation metric between transcript abundances for pairs of transcripts separated by genomic distance within each of multiple genomic distance bins; and generating a biomarker metric value derived from the correlation metrics for the multiple bins, wherein the biomarker metric characterizes the range of genomic distances in which pairs of transcripts have correlated expression, and samples having different chromatin states have different values ​​of the biomarker metric. Related methods, systems, and products are also described.

Description

Related applications

[0001] This application claims priority to a concurrently pending U.S. patent application filed on 12 June 2023, application number 63 / 507,720, entitled "CHROMATIN STRUCTURE BIOMARKERS." The subject matter of this related application is incorporated herein by reference. [Technical Field]

[0002] This disclosure relates to a method for characterizing a sample using biomarkers that exhibit chromatin structure derived from expression data. Related methods, systems, and products are also described. [Background technology]

[0003] The regulation of multiple genes and other transcriptable elements is known to be fundamental to all genomic function. Conventional methods such as bulk RNA-seq analysis provide data that can be useful for grouping or classifying genes (or samples) based on transcript abundance. This is typically achieved by using feature extraction and dimensionality reduction algorithms such as partial least squares (e.g., see Non-Patent Document 1), slice inverse regression (e.g., see Non-Patent Document 2), principal component analysis (PCA) (e.g., see Non-Patent Document 3), and more recently, word embeddings (e.g., see Non-Patent Document 4). These approaches can be useful for identifying affected phenotypes or for identifying genes that may be co-regulated and have similar functions. Another important technique is to construct a network of genes based on these expression patterns across multiple samples (e.g., see Non-Patent Document 5). Once the network is constructed, it can help identify groups of co-expressed genes and identify key regulatory genes or pathways that may drive the observed expression patterns. These methods have greatly improved our understanding of the co-expression sets of genes involved in shared biological pathways or processes.

[0004] The regulation of gene expression depends on the local chromatin environment of the gene (Non-Patent Literature 6). Genes located within dense heterochromatin regions generally exhibit lower levels of expression compared to genes located in more open euchromatin regions. In this regard, the chromatin environment constrains transcription. Heterochromatin is a more condensed, transcriptionally attenuated structure that is less accessible to RNA polymerase and other transcription factors. Euchromatin is more open, more accessible, and more readily transcribed. Epigenetic modifications, including DNA methylation and histone acetylation, play important roles in gene regulation and the formation of functionally and structurally distinct regions in heterochromatin and euchromatin. Changes in local chromatin structure due to these modifications can affect gene expression by regulating the accessibility of DNA to transcription factors (Non-Patent Literature 7). On a relatively small scale, rearrangement of nucleosomes near promoter regions can upregulate some genes and downregulate others in response to stress or changes in the cellular environment (Non-Patent Literature 8). In very recent studies, human lung cells and umbilical vein cells have been analyzed and found that senescent cells contain fewer nucleosomes, promoting the movement of RNA polymerase II (Pol II), which allows it to move more quickly. This suggests that structural changes to DNA can dramatically affect the transcription mechanism, leading to decreased accuracy and increased error as RNA copies are made (Non-Patent Literature 9). Overall, degradation of heterochromatin domains can cause widespread changes in genomic composition, activating previously silenced genes, leading to abnormal gene expression and genomic instability. These structural changes in heterochromatin are associated with aging and related genetic disorders (e.g., Hutchinson-Gilford progeria (HGPS) and Werner syndrome (WRN)) (Non-Patent Literature 10, Non-Patent Literature 11).

[0005] Nevertheless, the dependence of gene expression on chromatin structure or the local environment is largely unknown.

Prior Art Documents

Non-Patent Documents

[0006]

Non-Patent Document 1

Non-Patent Document 2

Non-Patent Document 3

Non-Patent Document 4

Non-Patent Document 5

Non-Patent Document 6

Non-Patent Document 7

Non-Patent Document 8

Non-Patent Document 9

Non-Patent Document 10

Non-Patent Document 11

Summary of the Invention

[0007] Considering that chromatin structure, by definition, creates different regions of DNA that are more open or more densely packed along the chromosome, and that this localized genomic packing affects transcription, the inventors assumed that gene expression should be spatially correlated, taking into account the relative positions of genes within these different regions. The inventors further assumed that the degree of correlation between adjacent genes should depend on the similarity of their chromatin structure and accessibility. The inventors further assumed that, if this is indeed the case, the abundance of transcripts measured by RNA-seq should be correlated along the chromosome, and that genes in euchromatin domains should be upregulated compared to genes in heterochromatin domains. The inventors further recognized that these euchromatin and heterochromatin domains do not need to be contiguous along the chromosome. Rather, the average length of these domains should be determined by the rate at which these correlations decay for genes with higher or lower expression, respectively.

[0008] The inventors further hypothesized that heterochromatin loss associated with aging-related disorders causes changes in chromatin structure, which in turn affects the correlation scale, and therefore this correlation scale should serve as a useful biomarker for classifying disease states from healthy states. Accordingly, the inventors embarked on developing an integrated, data-driven approach to test the hypothesis that epigenetic modifications resulting from aging or pre-aging states can alter chromatin structure and ultimately affect long-term correlations between gene expressions. Using a top-down approach, the inventors systematically explored pairwise correlations of transcript abundances between distant genes as a function of their proximity. This method was applied to various whole-genome (bulk RNA) transcriptome datasets from human fibroblasts to establish transcript-transcript correlations based on these abundances. The inventors then utilized correlation length as a feature for grouping cells into different states (healthy vs. diseased) and demonstrated that this metric can be practically used as a biomarker for diseases, disorders, and states affecting chromatin structure, such as aging-related disorders. In particular, in the context of age-related disorders, the inventors utilized correlation length as a feature to group cells from disease / aging samples treated with the test drug (rejuvenated), samples not treated with the test drug (disease / aging), and healthy samples into different states. Biological replicas of the same state clustered together, showing a clear pattern of being closer to biologically similar samples (e.g., rejuvenated samples being closer to healthy samples). This demonstrated that this metric can be practically used as a biomarker for the therapeutic effect of diseases, disorders, and conditions affecting chromatin structure, such as age-related disorders.

[0009] Thus, according to the first aspect, a method for determining the chromatin state of a sample, comprising: receiving, by a processor, transcript abundance data for a plurality of transcripts from the sample; determining, by the processor, for each of a plurality of genomic distance bins, a correlation metric between the transcript abundances for pairs of transcripts separated by the genomic distance within each bin; generating, by the processor, a value of a biomarker metric derived from the correlation metrics for the plurality of bins, wherein the biomarker metric characterizes a range of genomic distances for which pairs of transcripts have correlated expression, and samples having different chromatin states have different values of the biomarker metric. A method is provided that includes these steps.

[0010] As will be appreciated by those skilled in the art, the complexity of the operations described herein (at least due to the complexity of analyzing expression data at the genomic or chromosomal scale and the amount of data typically generated by sequencing transcriptome material) is such that they exceed the scope of mental activity. Thus, except where the context dictates otherwise (e.g., when sample preparation or acquisition steps are described), all steps of the methods described herein are computer-implemented.

[0011] Thus, according to this aspect, a computer-implemented method for determining the chromatin state of a sample, comprising: receiving transcript abundance data for a plurality of transcripts from the sample; determining, for each of a plurality of genomic distance bins, a correlation metric between the transcript abundances for pairs of transcripts separated by the genomic distance within each bin; generating a value of a biomarker metric derived from the correlation metrics for the plurality of bins, wherein the biomarker metric characterizes a range of genomic distances for which pairs of transcripts have correlated expression, and samples having different chromatin states have different values of the biomarker metric. A method is also described herein that includes these steps.

[0012] The method of this aspect may have one or more of the following features.

[0013] The transcript abundance data may include the normalized abundances of each of a plurality of transcripts. The normalized abundance of a transcript may be obtained by dividing the abundance of the transcript by the average abundance of all transcripts from the sample, whereby a scaled abundance may be obtained. Alternatively or in addition to this, the normalized abundance of a transcript may be obtained by dividing the abundance or scaled abundance of the transcript by the sum of the abundances or scaled abundances of all transcripts from the sample, whereby a fractional abundance may be obtained.

[0014] Each of the plurality of transcripts may be associated with a genomic locus in a reference genome such as the human reference genome. The sample may be a human sample. The sample may be a sample containing a plurality of cells, preferably human cells. The sample may be a blood or tissue sample, or a cell sample (e.g., cell line).

[0015] Multiple transcripts are high-abundance transcripts, which are obtained by identifying a first group and a second group of transcripts in transcript abundance data for multiple transcripts, wherein the first group of transcripts may be present in higher amounts in the sample than the second group of transcripts. The method may include the processor identifying a first group and a second group of transcripts using transcript abundance data, wherein the first group of transcripts is present in higher amounts in the sample than the second group of transcripts, and the processor selecting the first group of transcripts before determining the correlation metric for multiple bins. Identifying the first and second groups of transcripts may be performed using a data-driven method. The data-driven method may include a clustering method and a curve-characteristic-based algorithm applied to a data series containing ranked abundances of transcripts as a function of rank. The curve-characteristic-based algorithm may be a kneele algorithm. The clustering method may be selected from spectral clustering, k-means, k-medoid, and density-based spatial clustering of applications with noise (DBSCAN). The clustering method may be spectral clustering. The clustering method may be DBSCAN. Spectral clustering may be performed using a kernel selected from Laplacian affinity, radial basis functions, or nearest neighbor method. The data-driven method may include a first data-driven method used for outlier detection (e.g., DBSCAN) and a second data-driven method used to identify transcripts in first and second groups from a set of transcripts after outlier detection (e.g., spectral clustering). The first and second data-driven methods may be selected from clustering methods and curve-characteristic-based algorithms applied to a data series containing ranked abundances of transcripts as a function of rank.

[0016] The genome distance bins are based on the chromosomal location of transcripts, and transcripts within a bin may be associated with genes that are the same number of genes separated from each other. Each genome distance bin contains pairs of transcripts associated with genes that are separated by n genes on the same chromosome, where n is a natural number or a range of natural numbers, and n may differ from bin to bin. The genome distance bins may include a first bin containing pairs of transcripts associated with genes that are separated by n1 genes on the same chromosome, and a second bin containing pairs of transcripts associated with genes that are separated by n2 genes on the same chromosome. In this embodiment, n1 is 0 and n2 is not 0. The genome distance bins may include a first bin containing pairs of transcripts associated with genes that are separated by n1 genes on the same chromosome, and a second bin containing pairs of transcripts associated with genes that are separated by n2 genes on the same chromosome. n1 and n2 are non-overlapping natural numbers or a range of natural numbers. For example, n1 may be 0 (i.e., the genes are directly adjacent on the chromosome), and n2 may be 1 (i.e., the genes are separated by only one gene on the chromosome).

[0017] The correlation metric may be a nonlinear correlation coefficient. The correlation metric may be a distance correlation. The correlation metric for the bins is:

[0018]

number

[0019] It is calculated as follows: During the ceremony, (x,y) is a vector of transcript abundances for all possible pairs of transcripts in the bottle. V may be the covariance of the empirical distances for the transcripts in the bottle. The empirical distance is given by the following formula:

[0020]

number

[0021] This is calculated using the formula, where D may be a matrix of all central Euclidean pairwise distances between transcript abundances of transcripts in a bin, centered on the row mean and column mean.

[0022] The biomarker metric derived from the correlation metric may be a value derived from the autocorrelation of the correlation metric. The autocorrelation of the correlation metric may be calculated as a function of the lag in the same units as the units of genomic distance associated with the bin. The autocorrelation of the correlation metric may be calculated as a function of the lag of the number of distant genes on the chromosome. The biomarker metric derived from the correlation metric may be a value proportional to the lag at which the autocorrelation of the correlation metric falls below a given confidence limit. The biomarker metric derived from the correlation metric may be a value proportional to the lag at which the autocorrelation of the correlation metric no longer differs significantly from zero at a selected confidence level. The confidence limit or level may be selected between 90% and 99%. The confidence level or limit may be 90%, 91%, 92%, 93%, 94%, 95%, 96%, 97%, 98%, or 99%. The correlation metric may be the lag at which the autocorrelation of the correlation metric falls below a given confidence limit when multiplied by 1000. The correlation metric can be multiplied by 1000, resulting in a lag where the autocorrelation of the correlation metric no longer differs significantly from zero at the selected confidence level.

[0023] Multiple transcripts may be located on the same chromosome. Multiple transcripts may be located on multiple chromosomes, and correlation metrics and biomarker values ​​may be determined separately for each chromosome of the multiple chromosomes, or for one or more chromosomes of the multiple chromosomes, using transcripts located on each chromosome. Multiple transcripts may be located on multiple chromosomes, and the method may include the processor selecting multiple transcripts located on the same chromosome before the processor determines the correlation metrics. The processor's selection of multiple transcripts located on the same chromosome may include selecting multiple transcripts located at any location on the chromosome, or selecting multiple transcripts located on a portion of the chromosome. The sample may be a human sample. The chromosome may be chromosome 6 or chromosome 7. In particular, the chromosome may be chromosome 6. The method may further include repeating the step of selecting multiple transcripts located on the same chromosome for further chromosomes or multiple further chromosomes, and determining the correlation metrics and biomarker values ​​for further chromosomes or multiple further chromosomes. Further chromosomes or multiple chromosomes may include chromosome 6 and / or 7.

[0024] Transcriptor abundance data may be obtained from whole transcriptome RNA sequence data. Transcriptor abundance data may include abundance data from each transcript detectable in the sample using transcriptome-wide transcriptome analysis techniques such as RNA sequencing. Biomarker metrics may be generated from transcript abundance data from multiple transcripts located on the same chromosome, and the value of the biomarker metric may be scaled by the number of transcripts on the chromosome. Transcriptor abundance data may include data from multiple transcripts located on the same chromosome, and the generation of a biomarker metric value by the processor may include the processor scaling the generated value by the number of transcripts on the chromosome, and optionally determining the number of transcripts located on the chromosome.

[0025] The method may further include generating biomarker metric values ​​for each of a plurality of samples by the processor, and clustering the obtained values. The clustering may be performed using hierarchical clustering and Manhattan distance. The method may include determining inter-cluster distances by the processor based on biomarker metric values ​​for samples in at least one pair of clusters. The pairs of clusters may include a cluster containing normal samples and a cluster containing test samples. The method may include comparing samples in a first cluster with samples in a second cluster using inter-cluster distances between the first and second clusters, between the first and third clusters, and between the second and third clusters. The inter-cluster distance may be determined by calculating the distance between biomarker values ​​for each pair of samples, including samples from each of the pairs of clusters, thereby obtaining multiple distances and then obtaining an aggregated distance from the multiple distances. The distances may be Manhattan distances. The aggregated distance may be average distances.

[0026] The method may include obtaining transcript abundance data from a sample. Obtaining transcript abundance data from a sample may include processing RNA sequence data, including RNA sequencing reads, to determine the abundance of each of several transcripts based on the reads that map to each transcript among several reads. Obtaining transcript abundance data from a sample may include obtaining RNA sequencing reads from a sample by RNA sequencing. The transcript abundance data may be obtained from RNA sequence data. The RNA sequence data may include multiple sequencing reads. The sequence data may include multiple aligned sequencing reads, for example, in the form of a SAM or BAM file. The sequence data may include multiple sequencing reads, each associated with a coordinate in the reference genome or transcriptome where the sequencing reads are aligned.

[0027] The method may include obtaining RNA sequence data from a sample and processing the RNA sequence data to obtain transcript abundance data for multiple genes. The step of processing the RNA sequence data may include one or more of the following: selecting one or more RNA sequence reads based on one or more quality metrics; removing one or more transcripts based on one or more criteria regarding the number of reads that map to transcripts in one or more samples; removing batch effects; and normalizing the RNA sequence data.

[0028] The step of obtaining RNA sequence data from one or more samples derived from the target includes, or may include, receiving sequence data from a user (e.g., via a user interface), from one or more computing devices, or from one or more data stores or databases. The step of obtaining RNA sequence data may further include sequencing the sample (or determining the sequence composition of transcriptome material present in the sample).

[0029] The method may include identifying the genomic coordinates of multiple genes.

[0030] The method may further include, for example, providing the user with, via a user interface, one or more of the following: values ​​of a biomarker metric, information identifying the chromatin state derived from said values, values ​​of a correlation metric for each of one or more of a plurality of genomic distance bins, information derived from clustering of a plurality of samples, including samples based on the values ​​of the biomarker metric for each sample, and optionally one or more further metrics (e.g., clusters associated with one of the plurality of samples, distances between one or more clusters, etc.).

[0031] A second embodiment provides a method for observing the effect of one or more test agents on cellular senescence or disease, the method comprising optionally combining the test agent with cells; obtaining transcript abundance data from a sample of cells exposed to the test agent; determining the chromatin state of the sample using a method of any embodiment of the first embodiment; and comparing the biomarker value of the sample with one or more corresponding control values.

[0032] The method according to this embodiment may have any one or more of the following optional features.

[0033] Comparing the biomarker values ​​of a sample to one or more corresponding control values ​​may involve clustering the biomarker values ​​of the sample with biomarker values ​​determined for one or more control samples using the method of any embodiment of the first aspect. The control samples may include one or more samples selected from aged or diseased samples, healthy samples, and samples treated with cytogenetic agents.

[0034] A third embodiment provides a method for determining the age, rejuvenation, or reprogramming state of cells in a sample, comprising: obtaining transcript abundance data from the sample; determining the chromatin state of the sample using a method according to any embodiment of the first embodiment; and comparing the biomarker values ​​of the sample with one or more corresponding control values. The control values ​​may be obtained from one or more samples having known age, rejuvenation, or reprogramming state using a method according to any embodiment of the first embodiment.

[0035] According to a fourth aspect, a method is provided for treating a subject diagnosed with an age-related disorder, comprising administering a therapeutically effective amount of a drug identified as effective in treating the age-related disorder to the subject, wherein the drug has been identified as effective in treating the age-related disorder using a method according to any embodiment of the second aspect.

[0036] The method according to this embodiment may have any one or more of the following optional features.

[0037] The method further comprises identifying agents effective in treating age-related disorders using a method according to any embodiment of the second aspect, wherein one or more test agents may include agents identified as effective in treating age-related disorders.

[0038] The control values ​​may include biomarker metric values ​​obtained using the method of any embodiment of the first aspect for at least one healthy sample and at least one sample containing cells with aging-related disorders (e.g., cells not treated with the test drug).

[0039] Age-related disorders may be HGPS or WRN.

[0040] A fifth aspect provides a method for conducting a reprogramming experiment, comprising: subjecting a sample of cells to be reprogrammed to an in vitro reprogramming protocol (e.g., incubation with multiple rejuvenation factors according to a predetermined protocol such as OSKM); and performing a method of any embodiment of the third aspect. By comparison with a control value, the reprogramming state of cells at any particular point in time in the in vitro reprogramming protocol can be identified, thereby making it possible to select protocol parameters (e.g., collection date, amount and / or addition time of one or more rejuvenation factors, etc.) that result in cells having a desired reprogramming state. For example, different cells are known to respond differently to the same reprogramming protocol, making it labor-intensive and difficult to adapt the protocol to obtain cells in a desired reprogramming state. Additionally, it is difficult to guarantee that such adaptations are successful. The methods described herein provide a solution to this by providing a biomarker that indicates the underlying chromatin state of the reprogramming state and can be measured using a simple and commonly available data source (e.g., RNA sequencing). Therefore, the method may include selecting the time of cell collection, and / or the amount and / or timing of cell exposure to one or more reprogramming factors, using comparison with a control value.

[0041] In a further embodiment, one or more non-temporary computer-readable media are provided, which, when executed by one or more processors, include instructions causing one or more processors to perform steps of any of the methods described herein, such as the method according to any embodiment of the first, second, or third embodiment described above.

[0042] In a further embodiment, a computer program is provided which includes code, when executed on a computer, causes the computer to perform steps of any of the methods described herein, such as the method according to any embodiment of the first, second, or third embodiment described above. [Brief explanation of the drawing]

[0043] [Figure 1] This is a flowchart illustrating a schematic method for characterizing the samples described herein. [Figure 2] This is a flowchart illustrating a schematic method for performing intervention screening according to the embodiments of this disclosure. [Figure 3] This figure shows an embodiment of a system for characterizing a sample and / or performing intervention screening, according to one embodiment of the present disclosure. [Figure 4] This diagram schematically illustrates the concept of long-range correlations between genes belonging to high-abundance classes (HAT, high-abundance transcripts) and low-abundance classes (LAT, low-abundance transcripts), using red and green colors, respectively. Both high-abundance and low-abundance transcripts can exhibit strong correlations. [Figure 5] This figure shows the unbiased clustering procedure used to separate all transcripts into two groups on chromosome 1, namely HAT and LAT. The x-axis represents the order of genes after ranking them in descending order of abundance. [Figure 6] This figure shows the distance correlation coefficient measured based on the abundance of binned pairs of transcripts as a function of Δm, where pairs in each bin are separated by the Δm between them. Red represents HAT transcripts, and green represents LAT transcripts. The inset shows the same data but is magnified to a Δm range between 0 and 500. [Figure 7] This figure shows the autocorrelation function as a function of lag for HAT (red, left) and LAT (green, right) classes. The correlation length is defined as the point at which the transcript-transcript autocorrelation falls below the 95% confidence limit. [Figure 8] This figure shows the hierarchical clustering of the feature matrix LHAT (correlation length of HAT genes for each sample (row) and chromosome (column)). The data shows clear clustering of samples based on treatment and disease, consistent integration at the replication level, and final integration of samples. WT = wild type; HGPS SCR = Hutchinson-Guilford progeria (HGPS) treated with scrambled ASO; HGPS NT = untreated HGPS; HGPS L1 ASO = HGPS treated with Line-1 ASO; WRN SCR = Werner syndrome (WRN) treated with scrambled ASO; WRN NT = untreated WRN; and WRN L1 ASO = WRN treated with Line-1 ASO. See Della Valle et al (2022). [Figure 9A] This figure shows the pairwise cluster distances of the samples in Figure 8 relative to the wild type. The data indicates that the ASO (antisense oligonucleotide) treated samples are much closer to the wild type than these disease counterparts. WT = wild type; HGPS SCR = Hutchinson-Guilford progeria (HGPS) treated with scrambled ASO; HGPS NT = untreated HGPS; HGPS L1 ASO = HGPS treated with Line-1 ASO; WRN SCR = Werner syndrome (WRN) treated with scrambled ASO; WRN NT = untreated WRN; and WRN L1 ASO = WRN treated with Line-1 ASO. See Della Valle et al (2022). [Figure 9B]This figure shows the normalized correlation length (IHAT) measured for disease samples taken from chromosomes 1 and 6. The figure shows that disease samples exhibited higher correlation lengths than ASO-treated / wild-type samples, and this effect was particularly pronounced on chromosome 6. WT = wild-type; HGPS SCR = Hutchinson-Guilford progeria (HGPS) treated with scrambled ASO; HGPS NT = untreated HGPS; HGPS L1 ASO = HGPS treated with Line-1 ASO; WRN SCR = Werner syndrome (WRN) treated with scrambled ASO; WRN NT = untreated WRN; and WRN L1 ASO = WRN treated with Line-1 ASO. See Della Valle et al (2022). [Figure 10A] Figures 8 and 9 show a comparison of clustering using correlation length in scenarios that consider only the LAT gene in the dataset. [Figure 10B] Figures 8 and 9 show a comparison of clustering using correlation length in scenarios that consider only the HAT gene in the dataset. [Figure 10C] Figures 8 and 9 show a comparison of clustering using correlation length in scenarios that consider all genes in the dataset. [Figure 10D] Figures 8 and 9 show a comparison of clustering using correlation length in scenarios that consider only the HAT gene in the dataset but do not include chromosome 6. [Figure 10E] Figures 8 and 9 show a comparison of clustering using correlation length in scenarios that consider only the HAT gene in the dataset but do not include chromosome 1. [Figure 10F] Figures 8 and 9 show a comparison of clustering using correlation length in scenarios that consider the HAT gene multiplied by the normalized transcript abundance of PCA1 in the datasets. [Figure 11A]Figures 8-9 show the results of hierarchical clustering using PCA (Principal Component Analysis) loading or values ​​from SVD (Singular Value Decomposition) of expression levels in the datasets. Figure 11A shows the results of hierarchical clustering using only the first principal component (PC). The data shows that disease samples (NT and SCR) aggregate with each other rather than with each other's types. [Figure 11B] Figures 8 and 9 show the results of hierarchical clustering using PCA (Principal Component Analysis) loading or values ​​from SVD (Singular Value Decomposition) of expression levels in the datasets shown. Figure 11B shows the results of hierarchical clustering using the first 10 PCs. In this case as well, disease samples are not separated. [Figure 11C] Figures 8 and 9 show the results of hierarchical clustering using PCA (Principal Component Analysis) loading or values ​​from SVD (Singular Value Decomposition) of expression levels in the datasets shown. Figure 11C shows the results of hierarchical clustering using SVD. [Figure 11D] This figure shows the mean l* values ​​calculated for chromosome 6 for different sample groups in data from Della Valle et al. [Figure 12A] This figure shows the results of hierarchical clustering using correlation length for the dataset by Fleisher et al. (2018). [Figure 12B] This figure shows the results of hierarchical clustering using the first 10 PCs of normalized transcript abundance for the same dataset as Figure 12A. [Figure 12C] This figure shows a box plot of the mean l* across chromosomes for each age group in the Fleischer dataset. [Figure 12D] This figure shows the mean l across chromosomes for discrete ages in the Fleischer dataset, rather than age bins, using a fitted linear regression model (linear regression coefficient = 0.0173, p = 2.01769 × 10⁻¹⁰). [Figure 13A]This figure schematically shows the spatial distribution (along the genomic coordinates) of gene expression (bottom) associated with RNA polymerases (black circles) that transcribe DNA in healthy young samples with intact closed chromatin and intact histone density. [Figure 13B] This figure schematically shows the spatial distribution (along the genomic coordinates) of gene expression (bottom) associated with RNA polymerase (black circles) that transcribes DNA in aging samples accompanied by heterochromatin loss. [Figure 14A] This figure shows gene expression data as a function of genomic coordinates on chromosome 6 in WT cells, based on data from Dalla Valle et al. [Figure 14B] This figure shows gene expression data as a function of genomic coordinates on chromosome 6 in HGPS-untreated cells, based on data from Dalla Valle et al. [Figure 14C] This figure shows gene expression data as a function of genomic coordinates on chromosome 6 in ASO-treated cells, based on data from Dalla Valle et al. [Figure 15A] This figure shows gene expression data corresponding to genomic coordinates on chromosome 12 in a 51-year-old donor (data from Fleischer et al. 2019). [Figure 15B] This figure shows gene expression data corresponding to genomic coordinates on chromosome 12 in a 94-year-old donor (data from Fleischer et al. 2019). [Figure 16] This figure shows plots illustrating the relationship between autocorrelation (bottom plot) and variance for multiple single spatial distributions of gene expression with different variances (shown in the top plot). [Figure 17] This figure schematically illustrates the method according to the embodiments of the present disclosure. [Figure 18A] This figure shows the results of hierarchical clustering using l* on the reprogramming dataset from Gil et al. 2022. [Figure 18B]This figure shows the absolute l* (average across chromosomes) for different cellular states identified in Gil et al. 2022. [Figure 18C] Figures 18A and 18B show the mean l* on chromosome 7 for the same sample, based on its condition on day 13 (error bars indicate the standard deviation across the sample). [Figure 19A] Figure 19A shows the mean l* values ​​for chromosome 6 in a dataset of HUVEC cells (human umbilical vein endothelial cells) cultured in the presence of chemicals being tested for their effects on cell rejuvenation. [Figure 19B] This figure shows the mean l* values ​​for chromosome 6 from the same dataset as Figure 19A, but with data for compounds that have been shown to have aging effects based on gene expression and methylation data. [Figure 19C] This figure shows the distribution of mean l* in HUVEC cells cultured over multiple passages as an in vitro model of aging. Figure 19C shows data for the first cell line, HUVEC B. [Figure 19D] This figure shows the distribution of mean l* of HUVEC cells cultured over multiple passages as an in vitro model of aging. Figure 19D shows the data for the second cell line, HUVEC D. [Figure 20A] This figure shows hierarchical clustering results using the intrachromosomal correlation length metric (l*=lHAT) calculated for high-expression transcripts selected using spectral clustering with different gamma parameter values. The data are from age-separated liver samples from GTEx. Figure 20A shows the results for gamma=3.5. [Figure 20B]This figure shows hierarchical clustering results using the intrachromosomal correlation length metric (l*=lHAT), calculated for high-expression transcripts selected using spectral clustering with different gamma parameter values. The data are from age-separated liver samples from GTEx. Figure 20B shows the results for gamma=2.5. [Figure 20C] This figure shows hierarchical clustering results using the intrachromosomal correlation length metric (l*=lHAT) calculated for high-expression transcripts selected using spectral clustering with different gamma parameter values. The data are from age-separated liver samples from GTEx. Figure 20C shows the results for gamma=1. [Modes for carrying out the invention]

[0044] This disclosure provides a method for determining the chromatin state of a sample using transcript abundance data, such as data from bulk RNA-seq. Different chromatin states may refer to different amounts or proportions of open chromatin versus closed chromatin in one or more regions of the genome, or different lengths of open chromatin regions in one or more regions of the genome. Different chromatin states may be associated with disease, spontaneous aging, pathological aging, rejuvenation, and / or reprogramming. Thus, the method described herein can be used to determine the state of cells in terms of natural age, pathological age, rejuvenation state, reprogramming state, disease state, and the effect of any perturbation (e.g., drug treatment) on any of these. Traditionally, biologists have used RNA-seq data to analyze gene expression and have had to rely on other techniques such as ATAC-seq to understand changes in chromatin structure. ATAC-seq is an expensive procedure compared to the relatively inexpensive and more widespread bulk RNA-seq. The inventors recognized that changes in chromatin structure affect not only gene expression but also the number of genes that can express themselves. Certain genes may be silenced if they are part of heterochromatin, while certain genes in euchromatin regions are transcriptionally active. Therefore, the inventors hypothesized that if the average length of the correlation between genes in the transcriptional region could be captured, this should provide a measure of changes in chromatin structure for a particular sample (e.g., cell line). Because there are nonlinear relationships between genes, the inventors recognized that simply applying a metric such as Pearson correlation would not adequately capture this phenomenon. To "explain" the nonlinear relationships between genes, the genes are bucketed based on their relative distance from one another. Thus, bucket_k contains a list of gene pairs that are "k genes apart" on the chromosome. Nonlinear correlation metrics such as distance correlation (also known in this embodiment as Brown correlation) can be used to capture the correlations between pairs of genes in the bucket.Having described the nonlinear relationships between genes, the inventors propose developing a metric to capture the dependencies between gene buckets. In particular, the inventors used autocorrelation to measure the dependencies between gene buckets, i.e., the correlation of buckets containing genes separated "two genes apart" that depend on buckets containing genes separated "one gene apart". Similarly, the correlation of buckets containing genes separated "three genes apart" depends on buckets containing genes separated "two genes apart", and so on. Since each bucket advances beyond the previous bucket by one gene distance, significant lag measurements using cross-autocorrelation and 95% confidence interval envelopes indicate the maximum mean gene influence length. Finally, the inventors further recognized that not all genes are equally important in these analyses. Therefore, the inventors proposed a data-driven approach to separate signals from noise by identifying gene sets (HATs) that are more beneficial to the above processes.

[0045] The described method can be used, for example, to quantify similarities between different cellular states (e.g., disease vs. treatment vs. health), measure changes in cellular state due to interventions (e.g., test drugs, candidate therapeutics, etc.), more accurately determine the true age (cellular age) of a subject, and detect disorders affecting chromatin state (e.g., certain Mendelian disorders) at an early stage by comparing a sample from a subject suspected of having a disorder with a control value (e.g., a control sample, e.g., a metric obtained using the method described herein for a healthy sample).

[0046] In this disclosure, the following terms are used and are intended to be defined as set forth below.

[0047] As used herein, “sample” may be a cell or tissue sample, biosynthetic fluid, or extract (e.g., RNA extract obtained from a subject) from which transcriptome material can be obtained for transcriptome analysis such as RNA sequencing (also referred to herein as “RNAseq” or “RNA-seq”). A sample may be a cell, tissue, or biosynthetic fluid sample (e.g., biopsy material) obtained from a subject. Such samples may be referred to as “subject samples” or “individual samples.” In particular, a sample may be a sample containing skin and / or blood cells. Skin and / or blood cells may be selected from fibroblasts, keratinocytes, oral cells, endothelial cells, lymphoblastoid cells, and / or cells obtained from hemoderma, dermis, epidermis, or saliva. For the purpose of obtaining RNA sequence data (also referred herein as “transcript abundance data”), a “sample” as used herein may be a cell or tissue sample, or extract (e.g., RNA extract obtained from a subject) from which transcriptome material can be obtained. The sample may be newly obtained from the subject or may have been processed and / or stored prior to transcriptome analysis (e.g., frozen, fixed, or subjected to one or more purification, concentration, or extraction steps). The sample may be a cell or tissue culture sample. Thus, the sample described herein may refer to any type of sample containing cells or transcriptome material derived therefrom, whether it is derived from a biological sample obtained from a subject or from a sample obtained from, for example, a cell line. In embodiments, the sample is a sample obtained from a subject such as a human subject. The sample is preferably mammalian-derived (e.g., a mammalian cell sample or a sample from a mammalian subject, e.g., a cat, dog, horse, donkey, sheep, pig, goat, cow, mouse, rat, rabbit, or guinea pig), preferably human-derived (e.g., a human cell sample or a sample from a human subject).Furthermore, samples may be transported and / or stored, collection may be performed at a location away from the sequence data acquisition (e.g., sequencing) location, and / or any computer-implemented method steps described herein may be performed at a location away from the sample collection location and / or the sequence data acquisition (e.g., sequencing) location (for example, computer-implemented method steps may be performed by a network computer such as a “cloud” provider).

[0048] "Normal sample," "healthy sample," or "wild-type sample" refers to a sample that is not expected to be in a disease state, for example, a sample derived from a healthy subject. A control sample may be a normal sample. A control sample may be a sample containing cells from a specific known developmental stage, such as embryonic cells or adult or mature cells. A "test" or "affected" sample, or simply a sample characterized using the methods described herein (for example, by determining biomarker values ​​described herein and optionally comparing them to control values ​​or values ​​determined from one or more control samples), may be a sample containing affected cells, for example, a sample derived from an individual with a disease or disorder. The disease or disorder may be an age-related disorder. The disease or disorder may be a progeria syndrome. The disease or disorder may be selected from Hutchinson-Gilford progeria (HGPS) and Werner syndrome (WS). A test sample may be a sample containing cells from a specific developmental stage, for example, cells obtained from an embryo or adult or mature cell, or cells derived from an elderly person. A “test” sample may be a sample containing cells treated with one or more test agents, or cells derived from a subject treated with one or more test agents. A “control” sample may be a sample containing cells not treated with one or more test agents, or cells derived from a target not treated with one or more test agents. Test or control samples may be derived from a single cell. The test agent may be a reprogramming factor. The reprogramming factor may be selected from Oct3 / 4, Sox2, Klf4, and c-Myc (“Yamanaka factor” or “OSKM”). The test agent may be a radioactive substance, a compound and drug, an antibody, a protein or peptide, or an antisense compound (e.g., ASO or siRNA). The test agent may be a gene perturbation, such as a decrease or increase in CRISPR-mediated gene expression. Test samples may be derived from a perturb-seq library.

[0049] The terms “sequence data” or “transcript abundance data” refer to information indicating the presence of transcripts in a sample having a particular sequence. Such information may be obtained using sequencing techniques, such as RNA sequencing or sequencing of captured genomic loci (targeted or panel sequencing), or using array techniques, such as expression arrays or other molecular counting assays. In the context of this disclosure, sequence data is typically obtained by RNA sequencing, particularly by next-generation sequencing. Therefore, sequence data may include sequencing reads. Information such as the count of sequencing reads having a particular sequence or covering a particular genomic location may be obtained from the sequencing reads using methods known in the art. When non-digital techniques, such as array techniques, are used, sequence data may include signals (e.g., intensity values) indicating the number of sequences in a sample having a particular sequence, for example, by comparison with a suitable control. Sequence data may be mapped to a reference sequence, such as a reference genome or transcriptome, using methods known in the art. This may result in aligned sequencing reads, for example, in the form of SAM or BAM files. Therefore, sequencing reads or equivalent non-digital signals may be associated with specific genomic locations (where “genomic location” refers to a location within the reference genome to which the sequence data is mapped).

[0050] The compositions described herein may be pharmaceutical compositions further comprising pharmaceutically acceptable carriers, diluents, or excipients. The pharmaceutical compositions may optionally contain one or more further pharmaceutically active polypeptides and / or compounds. Such formulations may be, for example, in a form suitable for intravenous infusion.

[0051] As used herein, “treatment” means reducing, alleviating, or eliminating one or more symptoms of the disease being treated compared to the symptoms before treatment. “Prevention” (or “prophylaxis”) means delaying or preventing the onset of symptoms of a disease. Prevention may be absolute (i.e., no disease occurs), or it may be effective only in some individuals or for a limited time.

[0052] As used herein, the term “computer system” includes hardware, software, and data storage devices for realizing a system or performing methods according to the embodiments described above. For example, a computer system may comprise a central processing device (CPU) and / or graphics processing device (GPU), input means, output means, and data storage, which may be realized as one or more connected computing devices. A computer system may comprise a display, or a computing device having a display for providing a visual output display. Data storage may include RAM, disk drives, or other non-temporary computer-readable media. A computer system may comprise a plurality of computing devices connected by a network and capable of communicating with each other through that network. It is expressly assumed that a computer system may consist of or include a cloud computer.

[0053] As used herein, the term “computer-readable media” includes, but is not limited to, any non-temporary media or combinations of media that can be directly read and accessed by a computer or computer system. Media may include, but are not limited to, magnetic storage media such as floppy disks, hard disk storage media, and magnetic tapes; optical storage media such as optical disks or CD-ROMs; electrical storage media such as RAM, ROM, and flash memory; and hybrids and combinations of the above, such as magnetic / optical storage media.

[0054] Characterization of chromatin states This disclosure provides a method for determining the chromatin state of a sample. An exemplary method is described with reference to Figure 1.

[0055] The method may include an optional step 100 in which one or more samples are obtained, comprising transcriptome material (e.g., one or more cell or tissue samples). Optionally, the samples may be sequenced in step 120 to obtain at least RNA sequence data. RNA sequence data can be obtained by RNA sequencing. Alternatively, the sequence data may have been previously obtained and may be received from a user interface, computing device, or database.

[0056] In an optional step 140, transcript abundance data for multiple transcripts are obtained from RNA sequence data. This may include obtaining normalized and / or fractional transcript abundances. Normalized transcript abundances for a sample can be obtained by dividing the transcript abundance of each transcript by the average transcript abundance across multiple transcripts. Fractional transcript abundances for a sample can be obtained by dividing the (optionally normalized) transcript abundance for each transcript by the sum of the (optionally normalized) abundances of all transcripts across multiple transcripts.

[0057] In step 160, a first group of transcripts and a second group of transcripts are identified using transcript abundance data, where the first group of transcripts (HAT, high-abundance transcripts) are present in higher abundances in the sample than the second group of transcripts (LAT, low-abundance transcripts). This may be performed using any data-driven approach, such as using clustering (e.g., spectral clustering). This may further include selecting transcripts from the second group of transcripts.

[0058] In step 180, transcripts to be mapped to a first selected chromosome are selected. This may involve obtaining the genomic coordinates of each transcript by mapping the transcript sequence to a reference genome. Alternatively, the transcripts may already be associated with genomic coordinates, or the genomic coordinates of the transcripts may be obtained from a database (e.g., Ensembl).

[0059] In step 200, all possible pairs of transcripts are generated from the multiple transcripts selected in step 180 and separated into multiple genomic distance bins. Each genomic distance bin contains pairs of transcripts associated with genes that are separated by n genes on the same chromosome, where n is a natural number or a range of natural numbers, and n is different in each bin.

[0060] In step 220, for each bin obtained in step 200, the correlation metric between the transcript abundances of all transcript pairs is calculated. The correlation metric is:

[0061]

number

[0062] The distance correlation coefficient may be calculated as follows: During the ceremony, (x,y) is a vector of transcript abundances for all possible pairs of transcripts in the bottle. V is the covariance of the empirical distances for the transcripts in the bottle, and the empirical distances are,

[0063]

number

[0064] The formula is calculated as follows, where D may be a matrix of all central Euclidean pairwise distances between the transcript abundances of transcripts in the bin, centered on the row mean and column mean.

[0065] In step 240, the biomarker value is derived from the correlation metric calculated for multiple bins in step 220. This may involve calculating the autocorrelation of the correlation metric as a function of the lag of the number of gene portions on the chromosome (i.e., the unit used for the bins, where each bin is associated with a lag value, e.g., lag = 1, 2, 3, 4, 5, etc., up to the maximum distance with respect to the number of genes between any pair of transcripts on the chromosome). This may further involve determining a confidence interval around the autocorrelation metric at each lag value. The biomarker value may be the maximum lag value at which the autocorrelation metric does not differ significantly from zero, at the level of confidence associated with the confidence interval. This value may be further scaled by the total number of genes considered on the chromosome. This may help make the biomarker values ​​comparable across chromosomes. This is not necessary when the same chromosome is being compared across samples.

[0066] Steps 180–240 may be repeated for one or more additional chromosomes. Thus, biomarker values ​​can be obtained for each of multiple chromosomes.

[0067] Steps 140–240 may be repeated for one or more additional samples. In step 260, the biomarker values ​​for multiple samples (which may include multiple biomarker values, one for each of multiple chromosomes) are compared, for example, by clustering in the illustrated embodiment.

[0068] In step 280, the inter-cluster distance is obtained between pairs of clusters (e.g., a cluster containing test samples and a cluster containing normal samples). This may be done by calculating the Manhattan distance between biomarker values ​​for each pair of samples containing samples in each pair of clusters. The inter-cluster distance may be used to compare samples within two clusters to determine, for example, whether the treated samples are closer to normal samples than the untreated samples (by calculating the cluster distance between (i) the cluster of treated samples and the cluster of normal samples, and (ii) the cluster of untreated samples and the cluster of normal samples).

[0069] In the optional step 300, one or more results from the preceding steps are provided to the user, for example, through a user interface.

[0070] Application examples The methods described herein are applied in connection with perturbations and / or samples from disease subjects that are effective against chromatin structure, particularly, but not limited to, aging and age-related diseases and disorders. In fact, as shown in the following examples, the metrics described herein can be used to identify samples from genetic aging disorders from samples treated with therapies that have been shown to be effective in these diseases, and when characterized using the metrics described herein, it can indicate that the treated samples are closer to a healthy state. In other words, the methods described herein can identify rejuvenated samples from aging / disease samples before treatment and quantify their proximity to a healthy state. The metrics described herein can also be used to identify the cellular age of a sample, as demonstrated by using chronological age as a surrogate. Furthermore, the metrics described herein can also be used to diagnose diseases or disorders affecting chromatin structure, for example, by quantifying the metrics described herein on samples from subjects suspected of having such disorders and comparing these to metrics obtained using the methods described herein on one or more control samples.

[0071] Accordingly, this disclosure also provides a method for observing the effects of one or more test agents on aging or disease in human cells. This method may include combining a test agent with human cells (for example, for a specific time period such as at least one day, one week, or one month), and then observing the chromatin state of the human cells using the method described herein. The method then compares observations from cells exposed to the test agent with observations of the chromatin state of control cells not exposed to the test agent, or a control or reference value, determined using the method described herein, so that the effect of the test agent on aging or disease in human cells is observed. Optionally, the test agent is a compound, antibody, polypeptide, polynucleotide, antisense compound (e.g., ASO, siRNA) having a molecular weight of less than 3000, 2000, 1000, or 500 g / mol. The cells may be fibroblasts. The cells may be human cells.

[0072] Methods for observing the effects of one or more test agents on aging or disease in human cells using the methods described herein, and for comparing the effects of one or more test agents with those of different test agents, are also described herein. For example, the methods of this disclosure can be used to compare the effects of a cell neoplastic agent, such as OSKM, with those of a test agent to determine whether the test agent has similar properties to OSKM.

[0073] In some embodiments, the method includes administering one or more test agents to a human subject. In further embodiments, the method includes observing the effects of one or more test agents by isolating cells(s) or tissues from a human subject and then observing the effects of the test agents. Thus, methods for observing the effects of one or more test agents on aging or disease in a human subject are also described herein. The method may include administering the test agent to a subject and then observing the chromatin state of a sample containing cells derived from the subject using the method herein. The method then compares observations from cells derived from the subject treated with the test agent to observations of the chromatin state determined using the method herein for samples derived from the subject and / or from healthy subjects that have not been treated with the test agent, or for controls or reference values, so that the effects of the test agent on aging or disease in a human subject are observed.

[0074] A method for determining the cellular age of an individual is also described herein by comparing the values ​​of the biomarker metrics described herein, obtained from a sample of the individual's cells, with the corresponding values ​​obtained from a healthy sample having a known chronological age. The age of the individual may be determined as the age associated with one or more samples from healthy samples associated with the biomarker value closest to the biomarker value obtained for the individual. In other words, the age of an individual can be measured using the biomarkers described herein based on transcript abundance data from a sample of cells or tissue, for example, skin (e.g., fibroblasts) (or in other words, based on RNA extracted from the sample).

[0075] The biomarkers described herein are also useful in ex vivo studies of anti-aging interventions, thus enabling rapid evaluation of interventions based on real-time measurements of aging. Embodiments of the present invention are also useful in personalized medicine applications, as they enable the evaluation of accelerated or decelerated aging effects based on transcript abundance data (i.e., RNA expression data).

[0076] Also described herein are methods for observing biomarkers in a sample that correlates with an individual's age, such as a skin sample, and which include observing the individual's chromatin state by determining the values ​​of the biomarker metrics described herein. Since the biomarker metrics described herein are associated with an individual's age, determining the chromatin state of a sample described herein includes observing biomarkers associated with an individual's age. Typically, skin and blood cells are human fibroblasts, keratinocytes, buccal cells, endothelial cells, lymphoblastoid cells, and / or cells obtained from blood skin, dermis, epidermis, or saliva. Embodiments of this method further include using observations to estimate the individual's age (for example, by using regression analysis or by comparing the biomarkers with biomarker values ​​obtained for a control sample having a known age).

[0077] Accordingly, this disclosure provides a method for identifying agents that can treat chromatin-related disorders such as age-related disorders, and a method for treating subjects diagnosed with age-related disorders. An exemplary method is described with reference to Figure 2.

[0078] In an optional step 20, a sample containing cells with aging-related disorders is obtained (disease sample). Optionally, one or more samples containing healthy cells may be obtained (normal sample). In an optional step 22, cells in the disease sample are exposed to one or more test agents. In an optional step 24, transcriptome data is obtained from the cell samples exposed to each test agent, and optionally from disease samples and / or normal samples not exposed to any test agent. In step 26, transcript abundance data is obtained from the transcriptome data for each sample. In step 28, the chromatin state of each sample is determined using the method described herein, for example, by referring to Figure 1. In step 30, one or more test agents are compared based on the determined chromatin state. For example, a test agent that brings the biomarker values ​​of disease samples treated with the test agent closer to the biomarker values ​​of normal samples (compared to the initial distance between biomarker values ​​obtained in disease samples not treated with the test agent) is considered effective in treating aging-related disorders. In an optional step 32, the subjects are treated with the investigational drug identified in step 30 as effective in treating age-related disorders.

[0079] system Figure 3 shows one embodiment of a system according to the present disclosure for characterizing a sample and / or screening for one or more perturbations. The system comprises a computing device 1 having a processor 101 and computer-readable memory 102. In the shown embodiment, the computing device 1 also comprises a user interface 103, which is illustrated as a screen, but may include any other means of conveying information to the user, such as through audible or visual signals. The computing device 1 is communicably connected, for example, via a network 6, to a sequence data acquisition means 3, such as a sequencer, and / or one or more databases 2 that store sequence data. The one or more databases may additionally store other types of information that may be used by the computing device 1, such as reference sequences, parameters, etc. The computing device may be a smartphone, tablet, personal computer, or other computing device. The computing device is configured to implement the method described herein. In an alternative embodiment, the computing device 1 is configured to communicate with a remote computing device (not shown), the remote computing device itself is configured to implement the method described herein. In such a case, the remote computing device may also be configured to transmit the results of the method to the computing device. Communication between computing device 1 and the remote computing device may be via a wired or wireless connection, for example, via a local or public network such as the public internet or Wi-Fi. Sequence data acquisition means 3 may be wiredly connected to computing device 1 and / or database 2, or it may communicate via a wireless connection, for example, via network 6 as shown in the figure. The connection between computing device 1 and sequence data acquisition means 3 may be direct or indirect (for example, via a remote computer).The sequence data acquisition means 3 is configured to acquire sequence data from nucleic acid samples, such as RNA samples extracted from cell and / or tissue samples. In some embodiments, the sample may be subjected to one or more pretreatment steps such as RNA purification, fragmentation, library preparation, and targeted sequence capture (e.g., poly(A) capture, exon capture, and / or panel sequence capture). The sequence data acquisition means is preferably a next-generation sequencer. The sequence data acquisition means 3 can be directly or indirectly connected to one or more databases 2 that can store sequence data (raw data or partially processed data).

[0080] The following are presented as examples and should not be interpreted as limiting the scope of the claims.

[0081] Examples Introduction In this study, the inventors present and illustrate a novel method for characterizing samples and perturbations based on spatial transcript-transcript abundance correlations (i.e., correlations and changes in correlations along genomic coordinates).

[0082] In particular, the objective of the studies described in the following examples was to develop an integrated, data-driven approach to test the hypothesis that epigenetic modifications resulting from aging or pre-aging states can alter chromatin structure, ultimately affecting long-term correlations between gene expressions. The following studies systematically explore pairwise correlations of transcript abundances between distant genes as a function of their proximity. This method is then applied to various whole-genome (bulk RNA) transcriptome datasets derived from human fibroblasts to establish transcript correlations based on these abundances. A novel metric, referred to as "correlation length," is then developed and proposed as a feature for grouping cells into different states (health vs. disease).

[0083] The methods described herein provide a solution to the problem of characterizing the chromatin state of a sample using gene expression data.

[0084] In particular, the inventors embarked on developing novel RNA-seq biomarkers to probe chromatin structure. The biomarkers were intended to capture changes in chromatin (e.g., to provide information on the restoration of chromatin integrity in rejuvenation treatments and the effects of aging on chromatin, if aging is associated with heterochromatin loss), changes in gene expression (e.g., to provide information on the restoration of cellular function in rejuvenation treatments), and to provide metrics that can distinguish patterns across different cellular states (e.g., distinguishing between young vs. old, disease vs. healthy vs. treatment). This study is based on the hypothesis that if chromatin structure constrains transcription, there is heterochromatin loss with aging, and one of the effects of rejuvenation is its restoration, and therefore, devising metrics that are proxies of open chromatin length should be extremely useful for monitoring aging and rejuvenation.

[0085] Example 1 - Correlation length as a measure of co-expression In this example, the present inventors propose and explain the concept of correlation length as a method for characterizing a sample.

[0086] method Data. Two publicly available whole-genome bulk RNA-seq datasets from human fibroblasts were used for autocorrelation analysis.

[0087] The first dataset (Line-1 or L1), initially published by Della Valle et al., contains seven human fibroblast samples, each containing three biological replicas: wild-type (WT), Hutchinson-Guildford progeria (HGPS; sometimes referred to as progeria in these examples) cells treated with scrambled ASO (HGPS SCR), untreated HGPS cells (HGPS NT), HGPS cells treated with Line-1 ASO (HGPS L1 ASO), and Werner syndrome (WRN; sometimes referred to as "Werner" in these examples) cells, WRN cells treated with scrambled ASO (WRN SCR), untreated WRN cells (WRN NT), and WRN cells treated with Line-1 ASO (WRN L1 ASO). This is available on NCBI SRA (www.ncbi.nlm.nih.gov / sra) under accession numbers PRJNA704498 and GSE198675.

[0088] HGPS and WS are genetic disorders of premature aging characterized by disordered heterochromatin (Scaffid and Misteli, 2005; Osorio et al., 2011; Ocampo et al., 2016; Zhang et al., 2015). HGPS causes rapid aging in children, beginning as early as two years of age. Heart problems or stroke are the leading causes of death in most children with HGPS. The average life expectancy for children with HGPS is approximately 15 years. Individuals with Werner syndrome are characterized by rapid aging beginning in early adolescence or young adulthood and an increased risk of cancer. Signs and symptoms include shorter-than-average stature, thinning and graying hair, skin changes, thin arms and legs, voice changes, and abnormal facial features. Della Valle et al. (2022) demonstrated that treatment with anti-LINE 1 ASO can improve aging characteristics in cells derived from HGPS and WS patients, for example, by reducing DNA methylation age and restoring heterochromatin histone markings.

[0089] The second dataset, obtained from Fleischer et al., consists of human fibroblasts from 143 samples from human subjects of varying ages ranging from 1 to 94 years. This is available in the Gene Expression Omnibus under accession number GSE113957 (www.ncbi.nlm.nih.gov / geo / ). Of the 143 donors, 10 suffer from Hutchinson-Gilford progeria (HGPS), a disease associated with premature aging. These were removed from the dataset to avoid misleading analyses and conclusions.

[0090] These two datasets are particularly useful for testing hypotheses that identify correlations between large-scale changes in gene expression and modifications of chromatin structure induced by epigenetic changes, such as heterochromatin loss in aging and related conditions.

[0091] Transcriptome data analysis. Transcripts were annotated using Ensembl identifiers, and these genomic loci were mapped from the reference human genome GRCh38.p13 using the BioMart-Ensembl tool. Transcripts that were not expressed in any of the samples (i.e., their abundance values ​​were 0 across the entire dataset) were excluded. For the Fleischer dataset, we further subdivided the dataset based on age ranges, i.e., 0–20, 21–40, 41–60, 61–80, and 81–100, to demonstrate age-dependent correlation distance trends and infer several characteristic features.

[0092] For all subsequent analyses, including the calculation of correlation length, each sample in the dataset is normalized by its mean and divided by the sum to obtain the proportion of each transcript's contribution to the total abundance.

[0093] Identification of high-abundance and low-abundance transcripts. For each sample individually, transcripts are sorted and ranked based on their abundance values, with rank 1 assigned to the highest-abundance transcript and the highest rank (i.e., the total number of transcripts considered in that sample) assigned to the lowest-abundance transcript. By rearranging the samples, a smooth monotonic decay of transcript abundances is obtained (see Figure 5), which can then be separated into two classes: high-abundance transcripts (HAT) and low-abundance transcripts (LAT) using a spectral clustering algorithm with Laplacian affinity. This step is performed to remove noise resulting from sequencing depth errors (low read counts) across different samples. Note that clustering is performed on normalized abundance data, and the ranks are shown for illustrative purposes to illustrate noise in the data, which is highlighted by the deviation of ranked data from linear relationships on a log-log scale.

[0094] Alternative kernels for spectral clustering can be used, such as RBF (Radiating Basis Function) kernels or nearest neighbor kernels. Further alternative clustering methods, such as k-means or k-medoid, can also be used. In general, any data-driven method for separating LAT and HAT genes (or more generally, any data-driven method for separating a first group of data points / regions in a graph from a second group of data points / regions in a graph, e.g., the "Kneedle" algorithm for identifying knee / elbow points in a curve, see Satopaa et al. 2011) may be used. The inventors found that spectral clustering using Laplacian affinity in particular works well (by comparing clustering of samples from HAT-derived correlation lengths obtained using data clustered using different algorithms for identifying HAT / LAT transcripts - see Example 2).

[0095] Distance correlation. Considering the division of genes into HAT and LAT classes, the inventors quantify the position dependence of gene expression by calculating the mean autocorrelation function of pairwise transcript abundances from normalized RNA-seq data. For each sample and each chromosome, pairwise normalized transcript abundances are binned based on the chromosomal position of the genes in the pair, i.e., all gene pairs separated by only one gene (directly adjacent) are in bin 1, all gene pairs separated by only two genes are in bin 2, and so on. The distance between genes is denoted as Δm. Therefore, bin 1 has Δm=1, bin 2 has Δm=2, and so on. For example, considering the following hypothetical arrangement of genes A, B, C, D, E, F, G on a chromosome, the gene pairs with Δm=1 are:<A,B> ,<B,A> ,<B,C> ,<C,B> ,<C,D> ,<D,C> ,<D,E> ,<E,D> ,<E,F> ,<F,E> ,<F,G> ,<G,F> Therefore, the gene pairs with Δm=2 are<A,C> ,<C,A> ,<B,D> ,<D,B> ,<C,E> ,<E,C> ,<D,F> ,<F,D> ,<E,G> ,<G,E> And so on. Bin i is gene n i This includes the transcript abundances of all possible pairs, n i n is the total number of gene pairs segregated by Δm=i in the group of genes considered (e.g., all HAT genes on the chromosomes considered). i It changes and typically decreases as i (i.e., Δm) increases. The distance correlation coefficient C(Δm) for each bin is calculated using the following formula (similar to the one described in Szekely, Rizzo and Bakirov, 2007):

[0096]

number

[0097] During the ceremony, (x,y) is a vector of transcript abundances for all possible pairs of genes in the bin (i.e., all pairs of genes in the sample data where the genes are separated by Δm genes on the genome). V represents the covariance of the empirical distance, and the empirical distance can be described with respect to the central Euclidean distance D using the following formula:

[0098]

number

[0099] In the formula, D is a distance matrix containing all pairwise distances between normalized transcript levels of genes in a bin, i.e., all pairwise distances centered on the row mean and column mean (||Xj-Xk|| and ||Yj-Yj||, respectively, where ||.|| represents the Euclidean norm). For each bin, there exists a pair of vectors (X,Y). One matrix D is calculated for vector X, and another matrix D is calculated for vector Y. Distance correlation is a measure of the dependence between two pairs of random vectors, and the population distance correlation coefficient is zero only if the random vectors are independent and can capture both linear and nonlinear associations. In contrast, Pearson correlation detects only linear associations between two random variables.

[0100] Distance correlation is a useful metric for identifying correlations between two random variables when a nonlinear relationship exists between them. Because transcript abundance or gene expression is highly nonlinear (i.e., transcript abundance is a nonlinear function of gene location, i.e., genes are not necessarily expected to correlate linearly as a function of gene location distance), other measures such as Pearson correlation may not be sufficient to capture complex relationships.

[0101] Correlation Length. The inventors further perform autocorrelation of C(Δm) to measure the relationship between a variable and its own lag value. This provides a measure for obtaining a length scale of the correlation phenomenon. Performing autocorrelation is useful for capturing the scale of distance correlation decay up to the point of sharp increase in “high” (Δm). The inventors then measure the correlation length for each chromosome as the point where the transcript-transcript autocorrelation falls below the 95% confidence limit (shaded area in Figure 7). While the 95% confidence limit was used as a threshold commonly used to view the confidence level of the metric, it should be noted that the results described herein work similarly for other choices of confidence limits, such as any confidence level from 90% to 99%, for example. When calculating autocorrelation of white noise data, the correlation of all lags falls inside the envelope (95% confidence interval, i.e., the white noise process has zero autocorrelation for all values ​​of the lag, in other words, the white noise is continuously uncorrelated). In other words, with white noise data, the value of the data at one point in a series of points does not predict the values ​​at later points in the series. In contrast, with data where high (significantly different from zero) autocorrelation exists, the autocorrelation for at least some of the lag values ​​lies outside the confidence interval envelope (i.e., significantly different from zero). When plotting the envelope of the 95% confidence interval as a function of lag, the intersection point between the autocorrelation data series and the envelope (i.e., where the autocorrelation values ​​lie outside the envelope) determines the largest lag that still yields significant autocorrelation. k The confidence interval (CI) of the autocorrelation AC in ) is:

[0102]

number

[0103] It can be calculated as follows, in the formula, AC SE,k teeth,

[0104]

number

[0105] is the standard error of the autocorrelation at lag k, calculated as AC i is the autocorrelation estimate at lag i ≤ k, and N is the number of time steps (here gene lags) in the sample.

[0106] Finally, a two-dimensional matrix L samples ×M chr (number of samples × number of chromosomes) is constructed, and its coefficients are the correlation lengths of each sample and chromosome, scaled by the total number of genes on each chromosome. For the hierarchical clustering of samples shown in FIG. 8, the inventors utilized the correlation lengths obtained from transcripts in the highly abundant classes that resulted in the accurate separation of the samples. HAT and L LAT For the hierarchical clustering of samples shown in FIG. 8, the inventors utilized the correlation lengths obtained from transcripts in the highly abundant classes that resulted in the accurate separation of the samples.

[0107] result Firstly, the inventors evaluated whether RNA-seq expression values ​​alone (i.e., without considering chromosome position in the analysis) could be used as a biomarker for aging. Figure 11C shows the clustering results for SVD against the LINE-1 dataset. Figures 11A and 11B show the clustering results using one principal component and ten principal components, respectively. These results indicate that SVD exhibits many interleaved cell lines. Clustering using principal components shows clearer separation at a high level (e.g., WT and ASO clustering while all progerias cluster together and all WRNs cluster together), but alternating arrangements of NT and SCR lines still exist. Figure 12B shows the clustering results using the Fleischer dataset with ten principal components (covering similar variance). This indicates that severe misclustering exists when PCA loading is used only for gene expression data. Note that the data was bucketed into five age groups as shown to ensure sufficient samples for each age group. This indicates that conventional RNA-seq analysis techniques have limitations. While they are useful for analyzing the effects resulting from downstream expression levels, the inventors hypothesized that these limitations stem, at least in part, from the fact that they do not incorporate modeling and understanding the underlying causes of expression differences between samples. As a result, the interpretability in explaining observations is limited. The inventors hypothesized that a novel measure incorporating several elements of functional and structural causes, used in conjunction with these conventional approaches, should provide a better and more complete picture of the underlying dynamics.

[0108] Gene transcription is one of many factors that depend on age and cellular state. Figure 13A schematically shows RNA polymerase (black circles) transcribing DNA in a healthy young sample with intact closed chromatin and intact histone density. For a particular histone density and epigenetic state, this results in unique gene expression plotted against the gene locations on the plots in the panel below (the x-axis represents genomic coordinates, and the plots show the spatial distribution of gene expression, not histograms). As RNA polymerase traverses the DNA, a variety of more genes are transcribed, resulting in a series of different types of distributions of gene expression normalized along the genome. This can be in any form, but here it is illustrated as a normal distribution with different parameters at each location. Horizontal arrows indicate the variance (spread) of each distribution. Figure 13B shows the same locations for senescent cells, where we assume there is a loss of heterochromatin. In that case, this should result in an epigenetic state where there is more open chromatin, a relatively lower histone density, and polymerase can access and transcribe more functional DNA. This, in turn, opens up adjacent sites and activates repressed genes. Now more genes are expressed, which results in a flatter and broader distribution compared to healthier young cells. Here again, the shape of the distribution shown is irrelevant, and for explanatory purposes, with respect to senescent cells, we assumed that there should be a higher variance of gene expression, both overall and locally. In particular, we observed that in senescent cells, the number of activated genes increases, which results in a higher variance of gene expression with respect to genomic coordinates (i.e., negative kurtosis of gene expression along genomic coordinates in senescent cells compared to healthy / young cells - a flatter and broader probability distribution).

[0109] This is illustrated in the actual dataset in Figure 14 (the LINE-1 dataset from Della Valle et al.). Figure 14A shows data from WT cells, Figure 14B shows data from HGPS NT cells, and Figure 14C shows data from ASO-treated cells. All data shown focus on a single chromosome (chromosome 6). The scatter plot at the bottom of each figure is the original dataset, with the x-axis being genomic coordinates and the y-axis being normalized RNA-seq data. A mixture of Gaussian distributions was fitted to the RNA-seq data (using the Dirichlet process algorithm) to construct the contour plots shown on the scatter plots. The corresponding Gaussian distributions are shown in the top panel of each figure. The dashed lines are the distributions obtained from the mixture of individual Gaussian distributions shown by the solid lines. By examining these plots, it becomes clear (particularly in the contour plots) that there is a slight increase in the variance of the distribution for affected progeria conditions (a flatter shape) compared to WT and ASO-treated samples. ASO treatment brings the "fitted" distribution closer to the WT, but the mean and variance of each distribution are still quite different from the WT due to the uncertainty and quality of the fit. This motivated us to look for metrics other than spatial distribution variance that could capture this mechanism. Figure 15 shows similar data for chromosome 12 from donors of different ages in the Fleischer 2019 aging dataset (Figure 15A shows data for a 51-year-old donor, and Figure 15B shows data for a 94-year-old donor). This indicates that there is higher variance for aged samples compared to younger samples.

[0110] In summary, the inventors hypothesized that the breakdown of heterochromatin leads to an increase in the number of expressed genes. If the inventors consider that the breakdown of heterochromatin occurs due to aging, then the aging process should open up many adjacent regions and alter the shape of the gene expression distribution along the chromosomal coordinates, resulting in increased variance (lower kurtosis). The challenge in capturing this phenomenon is that we do not want to assume any kind of distribution in the first place, and even if we did, it is difficult to "reliably" measure the increase in variance (as described above) due to the complexity of the data, the presence of high noise, and the fact that the data is expected to be a mixture of different distributions. Therefore, the inventors set out to design a method that can provide an alternative method to capture this phenomenon, and ideally, a method that can be used as a macroscopic holistic measure that takes into account the overall change in the characteristic properties of the gene expression distribution.

[0111] The inventors first tested evidence for the location-dependent nature of gene expression by measuring pairwise correlations of transcript abundances. Figure 4 shows a graph of epigenetically similar (or belonging to the same chromatin environment) genes that show higher correlations. Since low-expression genes can also correlate, correlations should exist between low-abundance transcripts. The existence of such correlations between adjacent genes that are part of the same region should be evident from the correlation length between transcripts.

[0112] As described above, the inventors begin by mean-normalizing the RNA-seq data for all samples. Next, they separate the transcripts in each chromosome into two classes (high-abundance transcripts (HAT) and low-abundance transcripts (LAT)) based on their abundance values. Figure 5 shows the mean-normalized transcript abundances for chromosome 1 of wild-type (3 copies) samples on a log-log scale, in order of decreasing abundance from Line-1 data. The rank (x-axis) is based on the magnitude of the transcript abundance. The two classes, HAT and LAT, are shown in red and green, respectively. The data shows points for all three copies (i.e., there are actually three points for any rank, one for each copy).

[0113] Next, the inventors binned pairwise transcript abundances based on these chromosome locations, and pairs in each bin were separated by the number of Δm of transcripts between them (see Methods for further details). For HAT transcripts (red, data series increasing first), examining the spatial dependence of the correlation coefficient C(Δm) measured using distance correlation, we found a sharp decay between Δm=10 and 20 (see inset in Figure 6 - note the peak at Δm=0 where C(Δm)=1 is very close to the y-axis), fluctuating around the mean up to Δm=200, beyond which there are not enough gene pairs to reliably quantify the correlation. A similar pattern can be seen for LAT class transcripts in Figure 6 (main) (green, data series increasing at higher values ​​of Δm).

[0114] By definition, the distance correlation coefficient is a non-negative number, where a value of 0 signifies independence between pairs of transcripts. The correlation is expected to decrease to zero as Δm increases (i.e., gene expression between pairs of genes becomes less correlated as further separated pairs are considered). However, because the number of transcripts for high Δm is smaller, the correlation shows a sharp increase towards the end. To obtain a scale at which the distance correlation monotonically decays, we calculated the autocorrelation of C(Δm). The autocorrelation of a data series is the correlation between a time series and its time-lag version. Autocorrelation is usually described in the context of time series data, but here time is replaced by space. In particular, in this case, the lag is related to the distance of the number of genes. Autocorrelation shows the influence / correlation of points separated by various distances. A distribution of gentler slopes is associated with longer influences. By performing autocorrelation, we were able to identify the range of Δm at which the distance correlation decays to a point where it increases sharply. Figure 7 shows the spatial decay of the transcript-transcript autocorrelation function I (i.e., decay as a function of delay with respect to the number of genes) for chromosome 1 of WT samples (3 copies) for two classes (HAT on the left and LAT on the right). Note that a similar pattern is observed for all chromosomes. The inventors combined the concept of autocorrelation with the concept of confidence interval envelopes to find the maximum significant lag indicating the length of correlation of gene expression signals along genomic coordinates. As described above, the underlying hypothesis is that decreased heterochromatin / histone density (e.g., due to aging or disease) results in increased cotranscription, which can be identified as longer-length correlations, leading to the hypothesis that it is possible to correlate chromatin structural changes with the length of gene expression correlations. In other words, the inventors hypothesized that chromatin structural changes due to aging or disease should result in longer-length correlations of gene expression. The inventors define the point at which the transcript-transcript autocorrelation falls below the 95% confidence limit (the shaded area in Figure 7 showing the 95% confidence limit of the autocorrelation for each value of the gene lag) as the correlation length (l HAT and l LAT) is defined, and its unit is defined in relation to the number of genes. For example, 27 genes belonging to the HAT class correlate on average on chromosome 1, while 154 genes in the LAT class correlate. Thus, the data show that the abundance of transcripts spatially correlates along the chromosome for both HAT and LAT transcripts, with the former decaying more rapidly. This is likely due to the fact that the majority of transcripts in low-abundance classes have similar values ​​(often close to 0). This can be seen from Figure 6 (main), where relative differences can be observed not only along the y-axis (absolute value of correlation) but also along the x-axis. The length of the x-axis depends on the number of transcripts identified in the HAT and LAT classes. Because the range of the x-axis differs significantly across the two regions, the inset in Figure 6 shows the difference in correlation over the same range. Since the correlation length varies across all chromosomes, we defined a metric which is the correlation length scaled by the total number of genes on each chromosome. Then, for each sample, we defined a two-dimensional matrix L for the HAT and LAT classes, respectively. HAT and L LAT Construct the matrix. The matrix contains the scaled correlation lengths for each chromosome and sample for each gene class (either HAT or LAT) (i.e., the vector for each sample, containing the scaled correlation lengths for each of the multiple chromosomes).

[0115] Theoretical evidence for the concept of autocorrelation linked to spatial gene expression distribution variance is shown in Figure 16, which illustrates the relationship between autocorrelation and variance in a univariate scenario. This considers a simple example with a distribution centered at mean 500. All these curves have different variances to indicate repressed gene activation resulting from heterochromatin opening. By performing autocorrelation on these curves, we can observe an increasing trend in the maximum significant lag (or significance correlation length), indicated in the figure by a red arrow indicating the point where the autocorrelation function intersects the 95% confidence interval. In a multivariate scenario (multiple distributions for each gene), points in each distribution correlate differently, making it impossible to measure the autocorrelation length on a "cumulative" distribution. Therefore, we devised a method for aggregating different distributions while maintaining intra-distribution and inter-distribution characteristics. This was done by sweeping the multivariate distributions using pairs of genes with varying genomic distances between them, as shown in Figure 4. For example, delta=0 explains all possible pairs that are directly adjacent, delta=1 explains all pairs of genes that have one gene between them, and so on for all deltas up to n-1. After collecting all pairs of genes with different separation distances in the corresponding buckets, it is possible to calculate the distance correlation for each bucket. As explained above, the distance correlation function is similar to other correlation functions, except that it additionally captures nonlinear relationships. This distribution is then passed to the autocorrelation function to find the maximum significant lag (i.e., the value of the autocorrelation is calculated as a function of the lag and the value of the lag where the autocorrelation is no longer significant at the chosen confidence level is identified). The end-to-end pipeline is shown in Figure 17, which shows that for each chromosome individually, the RNA-seq data (gene expression as a function of genomic coordinates) is denoised, then binned (each bin contains genes separated by the same number of genes along the chromosome), then the distance correlation is calculated for each bin, and finally the autocorrelation of the distance correlation is obtained, from which the correlation length l* is determined and used as a proxy for the open chromatin length.In short, typical RNA-seq data passes through this pipeline to obtain the l* value per chromosome per sample. This is then normalized by the number of genes per chromosome. In some of the results shown below, this value is multiplied by 1000 to obtain a metric that has an integer range (and is therefore easy to handle) and can be interpreted as the intrachromosomal correlation length per 1000 genes. Example 2 demonstrates a systematic validation of this method.

[0116] Therefore, the data in this example demonstrate that a spatial correlation exists between the genes and the quantities of these transcripts.

[0117] Example 2 - Systematic validation in natural aging, pathological aging, reprogramming, and drug screening. In this embodiment, the method described in Example 1 is used to obtain metrics used to characterize the samples and to verify that these have the potential to identify differences between the samples and the treatments. In particular, this embodiment demonstrates the use of the algorithm described in Example 1 on two datasets - LINE-1 from Della Valle et al. and data from Fleischer et al. The objective here is, first, to verify whether l* has discriminative ability, and then to attempt to dig deeper into the discriminative drivers in order to understand what l* is capturing. Also, during this process, the inventors hoped that it would be possible to highlight the similarities and differences between spontaneous aging and pathological aging.

[0118] method Please refer to Example 1. As explained above, the correlation length (l*, also called intrachromosomal correlation length) is calculated using all transcripts, and the high-expression transcript (l HAT ) using or low-expression transcripts (l LAT ) can be calculated using ). Unless otherwise indicated, the results shown are for high-expression transcripts (l HAT It is obtained using the correlation length for ). Therefore, unless otherwise indicated, l* is l HATThis corresponds to the following. Figures 9, 10, and 12A show the results of the correlation length calculated without multiplying by 1000. All other figures showing the l* value use a scale in which the correlation length is multiplied by 1000.

[0119] Identification of high-abundance and low-abundance transcripts. For the data in Figures 8, 9, 10, and 12A, spectral clustering was used as described above. For the data in Figures 11D, 12C, 12D, and 18 onwards, in addition to spectral clustering, DBSCAN (Density-Based Spatial Clustering of Noisy Applications) was used to remove outliers before spectral clustering (denoising). DBSCAN can also be used instead of spectral clustering to identify high-abundance versus low-abundance transcripts. DBSCAN is advantageous because it is robust to variations in how RNA-seq data are processed, which is at least partially deterministic rather than probabilistic.

[0120] Sample clustering. To achieve optimal results, the inventors used correlation lengths obtained from transcripts in high-abundance classes, which resulted in accurate classification of the samples. Classification was performed using hierarchical clustering with mean-linked and Manhattan distance metrics. The inventors' approach utilized clustering to divide abundances into high and low classes (see Figure 5), resulting in improved clustering accuracy. However, when clustering was performed based on correlation lengths derived from all transcripts or low-abundance transcripts, the inventors found that the classification of disease and treatment samples from wild-type controls was suboptimal, as shown in Figures 10B and 10C. While it is true that excluding low-abundance transcripts from the analysis may result in the loss of important biological information (e.g., information encoding transcription factors or signaling molecules), limiting the inventors' analysis to high-abundance transcripts increased the accuracy and reliability of the inventors' clustering algorithm at the overall or chromosomal level. Furthermore, the inventors conducted extensive validation of the effects of clustering parameters on the results of the method. In particular, the spectral clustering described above was applied to data from GTEx (gtexportal.org / home / ) containing age-stratified samples from different tissues, using different values ​​of the parameter gamma, which affects the cutoff for selecting high-expression transcripts. Then, using the high-expression genes, l* was calculated as described above, and then these were used to cluster the samples using hierarchical clustering. The results are shown in Figure 20 for an exemplary tissue (liver here, Figure 20A shows the result for gamma=3.5, Figure 20B shows the result for gamma=2.5, and Figure 20C shows the result for gamma=1), indicating that the clustering is robust, as gamma values ​​from 3.5 to 1 still produced good clusters (and better separation than when PCA was applied to gene expression data, data not shown), although more interleaving of clusters may be observed as lower-expression transcripts are included (lower gamma).This demonstrates the robustness of the method and provides a way to evaluate the clustering parameters used to select highly expressed transcripts for l* calculation, enabling the selection of parameters that yield the most beneficial l*.

[0121] It should be noted that the steps described above were performed in addition to standard processing pipelines that included appropriate quality control, normalization, and statistical analysis, which the authors incorporated as preprocessing steps in their respective papers to improve the accuracy and reliability of the data.

[0122] To calculate the cluster distance used in Figure 9A, the inventors used the feature matrix L HAT Using the wild-type as a reference to compare the similarity between two experimental phenotypes, a cluster distance matrix (Cp) is constructed based on the Manhattan distance between each pair of samples. This is then averaged across chromosomes. For three copies (L1 dataset) in each phenotype, this becomes a 3x3 matrix (L for each chromosome for each pair of one treated sample (columns or rows) and one wild-type sample (columns or columns). HAT This yields the average distance between them, which can then be used to plot the average cluster distance along with error bars (standard deviations taken from the replica information). While other distance metrics are possible, it should be noted that the Manhattan distance has been found to work best. Although not bound by theory, this is thought to be because the Manhattan distance uses a range of values ​​more effectively than, for example, the Euclidean distance.

[0123] result To test whether the metrics developed in Example 1 have the potential to identify differences between samples or treatments, we use correlation lengths (experimentally and chromosomally) measured for HAT classes. Note that clustering performed based on correlation lengths obtained using all transcripts or low-abundance transcripts resulted in suboptimal clustering of samples, as will be further investigated below. The data suggest that LAT class transcripts can act as uniform background noise, and focusing only on high-abundance transcripts does not appear to diminish important signals. Figure 8 shows the feature matrix L where sample replicas are grouped together at the first level of merging. HAT This shows hierarchical clustering. It is noteworthy that in the next level of integration, "scrambled treatment" (i.e., treated with ASO having a scrambled sequence - negative control for ASO treatment) and "untreated" samples are integrated together, confirming the existence of similarities between diseased samples. This is because these samples were not treated with Line-1 antisense oligonucleotide (ASO) and can be associated with each other as controls for treatment. The clustering of LINE-1 ASO-treated samples and wild-type samples is consistent with experimental observations that LINE1 ASO can help bring diseased cells closer to the state of healthy cells (Della Valle et al., 2022). Depletion of L1 RNA from affected patients using ASO restored heterochromatin histone marks and reversed DNA methylation age, further demonstrating the potential modification of basal chromatin structure by the intervention (Della Valle et al., 2022).

[0124] In the previous section, the inventors considered the use of correlation length as a metric for separating cells in different states (e.g., diseased cells (e.g., those with premature aging disorders), treated cells, and samples of different ages). To ensure accurate analysis and avoid errors associated with sequencing depth, the inventors used a clustering algorithm for signal extraction. The inventors further evaluated the effectiveness of hierarchical clustering by analyzing the correlation lengths of genes in the HAT class, LAT class, and all transcripts, as shown in Figures 10A–10C, respectively. These results outline the importance of focusing on key targets contributing to classification and eliminating noise. Indeed, clustering using correlation length for HAT genes performed significantly better than using either LAT genes alone or all genes (both still performed to some extent, but the data indicate that removing LAT genes is beneficial for focusing on key informational signals). To further demonstrate the robustness of the method and highlight the significance of genome-wide analysis, the inventors performed sample clustering analysis after removing chromosome 6 (Figure 10D) and chromosome 1 (Figure 10E) from the analysis. These two chromosomes, due to their high density of histone genes, may significantly influence metrics characterizing chromatin structural changes resulting from damage or intervention.

[0125] The inventors have created a solution to complement existing workflows, as shown in Figure 10F, HAT We further propose that this can be integrated with P1 (the first principal component obtained by PCA of the complete transcript-level dataset). This integration results in relatively improved clustering performance compared to using PCA alone (see Figure 11A).

[0126] The inventors also compared their method to existing techniques by presenting clustering based on principal components of transcript abundance data. Figure 11A shows the clustering of samples using a feature matrix (P1) constructed using a first principal component of transcript abundance for each chromosome. The inventors observed that HGPS NT and HGPS SCR (untreated and scrambled treated) samples merge with each other rather than with their own types, which also apply to the disease Werner syndrome replicate. Having additional principal components is advantageous, as one principal component accounts for 44%–92% of the variance across all chromosomes. Figure 11B presents a more specific comparison with a feature matrix containing 10 principal components averaged across all chromosomes, covering approximately 99.9% of the variance, clearly showing that most of the disease state replicas do not merge sufficiently. Furthermore, the inventors observed that chromosome X affects the first principal component, and that removing it further worsens the clustering. This demonstrates that, through our approach, we can identify which chromosomes are most affected or important, as described above, but that the removal of these chromosomes does not adversely affect the output. Therefore, the data show that the proposed approach extracts patterns of structural changes across all chromosomes and is extremely robust. The Mann-Whitney U test did not identify a significant difference between the distribution of mean l* across chromosomes in WT and L1-ASO treated cells in the LINE-1 dataset. However, looking at individual chromosomes, we found a strong trend on chromosome 6 (see Figure 11D; the Mann-Whitney U test shows a significant difference between WT and L1-ASO vs. untreated (NT) and SCR), which indicates that pathological aging does not affect all chromosomes uniformly when l* captures chromatin structural changes. The inventors hypothesized that the importance of chromosome 6 must be related to the fact that HIST1, the largest cluster of histone genes (approximately 80%), is located on chromosome 6, and that chromosome 6 is also the house of the FOXO3 gene, which defines the hub of aging.The importance of chromosome 6 was also confirmed by gene expression analysis in the LINE-1 dataset. Among the top 200 differentially expressed genes identified by comparing disease versus WT (i.e., HGPS-NT versus WT and WRN-NT versus WT), chromosome 6 had the highest proportion of differentially expressed genes. This is also true overall (see Table 1). Therefore, pathological aging appears to induce various interchromosomal changes, with chromosome 6 being particularly prominent. The data in Figure 11D demonstrate that ASO treatment shows enhanced efficacy in aligning the cellular status of HGPS and WRN more closely with WT.

[0127] [Table 1]

[0128] Figure 12A highlights the hierarchical clustering for the Fleischer dataset. The Fleischer dataset is an aging dataset containing samples from neonates to elderly individuals (>90 years). The purpose of using this dataset is to demonstrate that the proposed measurement also works in the context of spontaneous aging. Furthermore, since a common underlying effect of pathological and spontaneous aging is chromatin changes (primarily loss of heterochromatin), this demonstrates that the proposed measurement acts as a "lens" to chromatin dynamics. However, since age in the Fleischer dataset is a continuous number, ages were bucketed to ensure that there are sufficient samples for each "category." 0-20, 21-40, etc., are arbitrary choices, and pre-processing such as 0-25, 26-50, etc., is also possible, as long as there are sufficient samples in each bucket. Due to this arbitrary bucketing, and the fact that categories reflect time series, there may be inconsistencies between clustering and age divisions (clustering is based on RNA-seq data that reveals cellular information). The data shows that the younger groups (ages 0-20 and 21-40) merge first. At the same time, on the other hand, we see an older population where the 61-80 age group merges with the 81 and over age group. It is worth noting that the existence of a small group of older populations that directly merges with the 61-80 age group indicates that the algorithm learns nuances (all of these samples are in their early 80s) and is not exclusively biased by age buckets. In other words, clustering shows a portion of the older population (age group 81-100) that merges with the 61-80 bucket. This group consists mainly of older adults in their early 80s who are more likely to be classified in the 61-80 bucket because they are cellularly younger and healthier. This verifies that the method does not derive a signal from how the data is age-bucketed; if so, this "cross-contamination" would not have occurred.

[0129] The middle-aged group (divided into 41-60 years old) representing the transitional state between the two subgroups is distinctive and ultimately merges due to its heterogeneity. Meanwhile, PCA clustering (Figure 12B) shows extremely poor performance; while the principal components capture variance, they fail to capture the underlying causes or the subtle nuances arising from them. Figure 12C shows a box plot of mean l* across chromosomes for each age group in the Fleischer dataset. This shows an overall increasing trend in l* across age buckets, which was shown to be statistically significant (Mann-Kendall test, trend = increasing, p = 0.027). Note that the inventors tested the use of mean l* across chromosomes and sum of l* across chromosomes in this dataset and other datasets, and found both to correlate strongly with spontaneous aging. Figure 12D shows the fitted linear regression model (linear regression coefficient = 0.0173, p = 2.01769 × 10⁻⁶). -10 The same data is shown for separate ages rather than age bins. This indicates that the metrics described herein increase significantly with age and can therefore be used as indicators of natural aging.

[0130] The data above demonstrate that l* exhibits sensitivity in detecting and capturing chromosomal structural changes. The data further indicate a significant variation in the DNA damage trajectory between spontaneous aging and pathological aging. In particular, spontaneous aging appears to affect all chromosomes, unlike pathological aging. The data further indicate that L1-treated ASOs appear to bring diseased cell lines closer to the wild type.

[0131] Building on this study on the LINE1 and Fleischer datasets to validate the l* metric, we set out to test the robustness of l* and see if we could gain more insights into partial reprogramming experiments using OSKM with a dataset from Gill et al., 2022. These reprogramming experiments used dermal fibroblasts from three middle-aged donors (38, 53, and 53) with epigenetic ages of 49, 45, and 55 (based on methylation patterns). The cells were treated with OSKM along with doxycycline for different lengths of time, and the cells were sorted on days 10, 13, 15, and 17. Those with SSEA4-positive, CD13-negative markers were labeled as “transient reprogramming intermediate,” and others were labeled as “transient reprogramming intermediate failure”: CD13-positive, SSEA4-negative. The authors also performed full reprogramming using the Sendai reprogramming protocol (this data was not used here). RNA-seq analysis was performed on these cells (see Figure 1e in Gill et al. 2022). This showed a restoration of the original identity of the old fibroblasts. In particular, Figure 1e in Gill et al. 2022 shows PCA plots performed on the whole reprogramming dataset + transient reprogramming, where all genes show a reprogramming trajectory along the first principal component PC1 from old fibroblasts to iPSCs, and other samples are clustered along this trajectory. Interestingly, transiently reprogrammed samples clustered at the start of this trajectory, indicating that these samples were again transcriptionally similar to fibroblasts, rather than reprogramming intermediates or iPSCs. Similar trajectories were also observed for the partial reprogramming dataset, but here five distinct states are clearly shown. We set out to analyze this data using the l* metric. Starting with clustering of samples based on the number of days of treatment, we hoped to see if l* could predict the transcriptional / epigenetic state of cells. The results are shown in Figure 18A, which demonstrates clustering using l*.This shows that transient reprogramming intermediates and iPSCs cluster together for all days of treatment, which makes sense as they are the closest. Additionally, cells and negative control intermediates, which are intermediates that cannot be transiently reprogrammed, also cluster together, which also makes sense as they represent their own states in Gill et al. 2022. Next, we plotted absolute l* (mean value across chromosomes) for the different states (in Figure 18B, the states are sorted based on day 13, which is the optimal day of treatment found by Gill et al.). This shows that the transiently reprogrammed state is the most closed state, iPSCs are the most open state, transient reprogramming intermediates are the next most open state, and fibroblasts are somewhere in between the two. The behavior shown in Figure 18B can be understood in a simple diagram (or model) of how euchromatin states change over time. This process begins with iPSCs having a lot of euchromatin, and as the differentiation process progresses, heterochromatin slowly begins to appear and euchromatin decreases. Subsequently, due to heterochromatin loss that occurs during aging, it returns to normal. Looking at individual chromosomes (data not shown), some chromosomes have higher l* than iPSCs, some have lower l*, but on average, iPSCs have higher l* compared to any other state. By performing differential gene expression analysis, it is possible to identify the most important chromosomes with the most differentially expressed genes and then observe the variation in l* on those chromosomes. The results are shown in Table 2, which shows that chromosome 7 is the most prominent relative to transiently reprogrammed iPSCs and the second closest relative to fibroblasts vs iPSCs. Chromosome 7 contains the elastin gene, which encodes elastin, the main component of elastic fibers, the main component of the extracellular matrix. Figure 18C shows the l* on chromosome 7 by state at day 13.This indicates that the l* metric strongly captured the expected trend, namely iPSCs and transient intermediates were likely the most open and therefore had the largest l*, while transiently reprogrammed ones that would have recovered their heterochromatin after OSKM had the lowest l*, and Fib was somewhere in between. This suggests that l* can capture different states in OSKM data based on underlying epigenetic changes to the chromatin.

[0132] [Table 2]

[0133] Next, the inventors undertook to investigate whether the novel metric could be useful in the context of chemical screening performed on HUVEC cells. The screening tested 18 chemicals and 1 control (DMSO vehicle) in two cell lines (HUVEC B and HUVEC D) (a total of 156 samples, including multiple biological copies per condition). The 18 chemicals were carefully selected across different categories, including oxidative stress / inflammation, protein homeostasis, epigenetic modifiers, and kinase / protease inhibitors. Cells were exposed to the chemicals from day 0 to day 3, then allowed to recover until day 7, at which point they were harvested for RNA-seq. The similarity distance between each treatment and DMSO was calculated separately for each cell line, using gene expression and l*. The same two compounds were found to deviate the most from DMSO in both cell lines, using both gene expression and l*. Analysis of methylation data for these two compounds using three different methylation clocks (AgeSkinBloodClock, DNAmAgingClock, and DNAmAgeHannumClock) showed that compound A significantly rejuvenated the clocks in all three clocks in HUVEC B and in two of the three clocks in HUVEC D. For compound B, the rejuvenation was not as pronounced as with compound A and was only observed in the DNAmAge HUVEC B panel.

[0134] Next, the inventors examined l* in both treatments compared to DMSO. The data in Figure 19A show significant changes in l* values ​​in chromosome 6 induced by the chemicals. On chromosome 6, compounds A and B significantly reduced l* values ​​in HUVEC B, indicating heterochromatin restoration. No effect on chromosome 6 was observed in HUVEC D. Compound A was also independently tested using methylation and gene expression changes to confirm its rejuvenating effect. This showed that compound A reduced H3K27me3 and significantly rescued aging-related transcriptome changes (reducing the log-fold changes in expression observed with aging). In contrast, another compound, compound C, significantly worsened these aging-related transcriptome changes. By examining the l* values ​​of samples treated with compound C, the inventors found that on chromosome 6, compound C increased the l* value in both HUVEC B and HUVEC D cell lines compared to the DMSO sample (statistically significant, p<0.05, Mann-Whitney U test in HUVEC B) (Figure 19B). Thus, at the individual chromosome level, chromosome 6 showed significant differences for all three chemicals introduced here, in the expected direction, with compounds A and B decreasing the l* value and compound C increasing the l* value compared to DMSO, at least in the HUVEC B cell line. The inventors then proceeded to further investigate the differences in behavior between the two cell lines.

[0135] The chemical screening dataset used includes various HUVEC cells undergoing passage to mimic the effects of aging. This can be used to investigate the in vitro aging effects in HUVEC cells. Figures 19C and 19D show the distribution of l* (mean l* and chromosome-wide variation) for each passage from young to old, 5 to 21 times, for HUVEC B cells and HUVEC D cells, respectively. Similar to the Fleischer dataset, Figure 19C shows a significant trend of increasing l* in HUVEC B cells that passed the Mann-Kendall test, reaching a plateau when the cells become relatively old (passage 17 and beyond). However, this is not observed in HUVEC D cells, and there is no systematic trend in l*. We believe this may provide an explanation for why chemical intervention is more pronounced in HUVEC B cells than in HUVEC D cells. This data shows that HUVEC B and HUVEC D cells exhibit different patterns, indicating the existence of a donor effect on chromatin structure associated with aging and the effect of chemicals on this. Nevertheless, the data also show that drugs effective against rejuvenation can be identified using the approaches described herein, as verified by orthogonal (methylation and gene expression-based) assays. The data also show that chromosome 6 appears to be a key chromosome in this process.

[0136] conclusion The studies described above reveal the existence of long-range spatial correlations between genes based on fundamental changes in chromatin structure due to epigenetic modifications. From an enzymatic perspective, the process of gene expression involves the binding of RNA polymerase to chromatin segments that must be accessible before this recruitment occurs. Changes in chromatin accessibility mediated by histone modifications and DNA methylation caused by disease or treatment alter the regulation of gene expression. Despite the richness and complexity of the underlying mechanisms of gene regulation, the studies described herein demonstrate that signatures of correlations between gene expressions based on these chromosomal locations can be formally quantified using distance correlations of pairwise transcript abundances. Correlation lengths measured across different chromosomes and samples have proven to be useful measures for separating healthy and diseased states, as well as different aging states, and as useful markers of rejuvenation and reprogramming, which have been shown to be more robust than conventional methods such as PCA. This analysis suggests the existence of long-scale correlations between gene expression driven by complex interactions between local changes in chromatin structure and gene activity.

[0137] References TIFF2026522340000012.tif101164

[0138] TIFF2026522340000013.tif240164

[0139] TIFF2026522340000014.tif117164

[0140] All references cited herein are incorporated herein by reference, in whole and for all purposes, to the same extent as each individual publication or patent or patent application is specifically and individually indicated so as to be incorporated as a whole by reference.

[0141] The specific embodiments described herein are provided as examples, not as limitations. Various modifications and variations of the compositions, methods, and uses described in this Art will be apparent to those skilled in the art without departing from the scope and spirit of the described Art. Any subtitles in this Specified Art are included for convenience only and should not be construed as limiting this Disclosure. Unless otherwise intended by context, the above descriptions and definitions of features apply equally to all described aspects and embodiments, and are not limited to any particular aspect or embodiment of the Invention.

[0142] Any method of any embodiment described herein may be provided as a computer program, or as a computer program product or computer-readable medium that carries a computer program configured to perform the above-described method when executed on a computer.

[0143] Throughout this specification and the claims, the following terms have the meanings expressly associated herein unless otherwise clearly indicated by the context. The phrase “in one embodiment” as used herein may refer to the same embodiment, though not necessarily the same embodiment. Furthermore, the phrase “in another embodiment” as used herein may refer to a different embodiment, though not necessarily a different embodiment. Thus, various embodiments of the present invention can be readily combined without departing from the scope or spirit of the invention, as described below.

[0144] Where used herein and in the appended claims, the singular forms “a,” “an,” and “the” refer to multiple subjects unless the context clearly indicates otherwise. Ranges may be expressed herein as “about” one particular value to and / or “about” another particular value. Where such ranges are expressed, another embodiment includes one particular value to and / or other particular values. Similarly, where a value is expressed as an approximation by the use of the antecedent “about,” it will be understood that a particular value forms another embodiment. The term “about” with respect to numbers is optional and means, for example, + / - 10%.

[0145] Throughout this Spec., including the following claims, unless the context requires otherwise, the words “comprise” and “include,” as well as variations such as “comprises,” “comprising,” and “including,” will be understood to mean that they include the integer or step or group of integers or steps described, but not that they exclude any other integer or step or group of integers or steps.

[0146] Other aspects and embodiments of the present invention provide the above aspects and embodiments in which the term “comprising” is replaced with the terms “consisting of” or “consisting essentially of,” unless the context should otherwise interpret them as such.

[0147] As used herein, “and / or” should be interpreted as a specific disclosure of each of two designated features or components, with or without the other. For example, “A and / or B” should be interpreted as if each of (i) A, (ii) B, and (iii) A and B were individually described herein.

[0148] Features disclosed in the above description, or in the following claims, or in the accompanying drawings, and expressed in these particular forms, or with respect to means for performing the disclosed functions, or methods or processes for obtaining the disclosed results, may be used, at their discretion, separately, or in any combination of such features, to implement the invention in its various forms.

Claims

1. A method for determining the chromatin state of a sample, The process involves a processor receiving data on the abundance of multiple transcripts from the sample, The processor performs the steps of determining, for each of a plurality of genome distance bins, a correlation metric between transcript abundances for pairs of transcripts separated by genome distance within each bin, A step of generating a biomarker metric value derived from the correlation metric for a plurality of bins using the processor, wherein the biomarker metric characterizes the range of genomic distances in which pairs of transcripts have correlated expression, and samples having different chromatin states have different values ​​of the biomarker metric. Methods that include...

2. The method according to claim 1, wherein the transcript abundance data includes the normalized abundance of each of the plurality of transcripts, and optionally, the normalized abundance of a transcript is obtained by dividing the abundance of a transcript by the average abundance of all transcripts from the sample, thereby obtaining a scaled abundance, and / or the normalized abundance of a transcript is obtained by dividing the abundance or scaled abundance of the transcript by the sum of the abundance or scaled abundance of all transcripts from the sample, thereby obtaining a fraction abundance.

3. The method according to claim 1 or 2, wherein each of the plurality of transcripts is associated with a reference genome, or optionally with a genomic locus in a human reference genome.

4. The plurality of transcripts are high-abundance transcripts, which are obtained by identifying a first group of transcripts and a second group of transcripts in the transcript abundance data for the plurality of transcripts, wherein the first group of transcripts has a higher abundance in the sample than the second group of transcripts, and / or The method includes the steps of: using the processor to identify a first group of transcripts and a second group of transcripts using the transcript abundance data, wherein the first group of transcripts has a higher abundance in the sample than the second group of transcripts; and selecting the first group of transcripts before determining the correlation metric for the plurality of bins, The method according to any one of claims 1 to 3.

5. The method according to claim 4, wherein the step of identifying first and second groups of transcripts is performed using a data-driven method.

6. The method according to claim 5, wherein the data-driven method comprises a clustering method and a curve-characteristic-based algorithm applied to a data series including ranked abundances of transcripts as a function of rank, wherein optionally the curve-characteristic-based algorithm is a Kneedle algorithm, and / or the data-driven method comprises a first data-driven method used for outlier detection and a second data-driven method used to identify first and second groups of transcripts from a set of transcripts obtained as a result of outlier detection, wherein optionally the first and / or second data-driven method is a clustering method.

7. The clustering method is selected from spectral clustering, k-means clustering, k-medoid clustering, and density-based spatial clustering of applications with noise (DBSCAN), and optionally, spectral clustering is performed using a kernel selected from Laplacian affinity, radial basis functions, or nearest neighbor clustering, according to claim 6.

8. The method according to any one of claims 1 to 7, wherein the genome distance bins are based on the chromosomal location of the transcripts, and the transcripts in the bins are associated with genes that are the same number of genes that are far apart from each other.

9. The method according to any one of claims 1 to 8, wherein each genome distance bin includes a pair of transcripts associated with genes that are separated by n genes on the same chromosome, where n is a natural number or a range of natural numbers, and n is different for each bin.

10. The method according to claim 9, wherein the genome distance bin comprises a first bin containing pairs of transcripts associated with genes separated by n1 genes on the same chromosome, and a second bin containing pairs of transcripts associated with genes separated by n2 genes on the same chromosome, wherein n1 is optionally 0 and n2 is not 0.

11. The method according to any one of claims 1 to 10, wherein the correlation metric is a nonlinear correlation coefficient, and optionally, the correlation metric is a distance correlation.

12. The correlation metric for the bins is: [Math 1] It is calculated as follows: During the ceremony, (x, y) is a vector of transcript abundances for all possible pairs of transcripts in the bottle, V is the covariance of the empirical distances for the transcripts in the bottle. The method according to claim 11.

13. The empirical distance is given by the following formula [Math 2] Calculated using the formula, where D is a matrix of all central Euclidean pairwise distances between the transcript abundances of the transcripts in the bin, centered on the row mean and column mean. The method according to claim 12.

14. The method according to any one of claims 1 to 13, wherein the biomarker metric derived from the correlation metric is a value derived from the autocorrelation of the correlation metric.

15. The method according to claim 14, wherein the autocorrelation of the correlation metric is calculated as a function of a lag in the same units as the units of genomic distance associated with the bin.

16. The method according to claim 14 or 15, wherein the autocorrelation of the correlation metric is calculated as a function of the lag of the number of distant genes on the chromosome.

17. The biomarker metric derived from the correlation metric is a value proportional to the lag at which the autocorrelation of the correlation metric falls below a predetermined confidence limit, and / or The biomarker metric derived from the correlation metric is a value proportional to the lag at which the autocorrelation of the correlation metric no longer differs significantly from zero at the selected confidence level. Optionally, the confidence limit or level is selected between 90% and 99%, and / or the confidence level or limit is 90%, 91%, 92%, 93%, 94%, 95%, 96%, 97%, 98%, or 99%, and / or the correlation metric is multiplied by 1000 in any case, resulting in a lag where the autocorrelation of the correlation metric falls below a predetermined confidence limit, and / or a lag where the autocorrelation of the correlation metric no longer differs significantly from zero at the selected confidence level. The method according to any one of claims 14 to 16.

18. The method according to any one of claims 1 to 17, wherein the plurality of transcripts are located on the same chromosome, or the plurality of transcripts are located on multiple chromosomes, and the correlation metric and biomarker value are determined separately for each chromosome of the multiple chromosomes using the transcript located on each of the multiple chromosomes.

19. The method according to any one of claims 1 to 18, wherein the plurality of transcripts are located on a plurality of chromosomes, and the method comprises the step of selecting a plurality of transcripts located on the same chromosome by the processor before the processor determines the correlation metric, optionally the method comprises the step of repeating the step of selecting a plurality of transcripts located on the same chromosome for a further chromosome or a plurality of further chromosomes, and the step of determining the correlation metric and biomarker values ​​for the further chromosome or a plurality of further chromosomes, optionally the sample is a human sample, and the chromosome or further chromosome is chromosome 6 or chromosome 7.

20. The method according to any one of claims 1 to 19, wherein the transcript abundance data is obtained from whole transcriptome RNA sequence data, and / or the transcript abundance data includes abundance data from each transcript detectable in the sample using transcriptome-wide transcriptome analysis techniques such as RNA sequencing.

21. The method according to any one of claims 1 to 20, wherein the biomarker metric is generated from transcript abundance data from a plurality of transcripts located on the same chromosome, the value of the biomarker metric is scaled by the number of transcripts on the chromosome, and / or the transcript abundance data includes data from a plurality of transcripts located on the same chromosome, and the step of the processor generating the value of the biomarker metric includes the step of the processor scaling the generated value by the number of transcripts on the chromosome, and optionally, the step of determining the number of transcripts located on the chromosome.

22. The method according to any one of claims 1 to 21, further comprising the steps of: generating a value of the biomarker metric for each of a plurality of samples using the processor; and clustering the obtained values.

23. The method according to claim 22, wherein the clustering step is performed using hierarchical clustering and Manhattan distance.

24. The method according to claim 22 or 23, further comprising the step of determining an inter-cluster distance based on the value of the biomarker metric for samples in at least one pair of clusters, wherein the processor further comprises the step of determining an inter-cluster distance based on the value of the biomarker metric for samples in at least one pair of clusters, wherein the pair of clusters optionally comprises a cluster containing normal samples and a cluster containing test samples, and / or the method comprises the step of comparing a sample in a first cluster with a sample in a second cluster using the inter-cluster distance between the first cluster and the second cluster, or between the first cluster and a third cluster, and between the second cluster and the third cluster.

25. The method according to claim 24, wherein the inter-cluster distance is determined by calculating the distance between biomarker values ​​for each pair of samples, including the samples from each of the cluster pairs, for each pair of samples, thereby obtaining a plurality of distances, and obtaining a combined distance from the plurality of distances, wherein optionally the distance is the Manhattan distance and / or the combined distance is the average distance.

26. The method according to any one of claims 1 to 25, wherein the method includes the step of obtaining transcript abundance data from the sample, the step of obtaining transcript abundance data from the sample includes processing RNA sequence data including RNA sequencing reads to determine the abundance of each of the plurality of transcripts based on the reads that map to each of the plurality of transcripts from among the plurality of reads, and / or the step of obtaining transcript abundance data from the sample includes the step of obtaining RNA sequencing reads from the sample by RNA sequencing.

27. A method for observing the effect of one or more test agents on cellular aging or disease, The steps include combining the aforementioned test drug with cells, The steps include obtaining transcript abundance data from the cell sample, A step of determining the chromatin state of the sample using the method described in any one of claims 1 to 26, The steps include comparing the biomarker value of the sample with one or more corresponding control values, Methods that include...

28. The method according to claim 27, wherein the step of comparing the biomarker value of the sample with one or more corresponding control values ​​includes the step of clustering the biomarker value of the sample with biomarker values ​​determined for one or more control samples using the method according to any one of claims 1 to 25.

29. The method according to claim 27 or 28, wherein the control sample comprises one or more samples selected from aged or diseased samples, healthy samples, and samples treated with a cell neogenetic agent.

30. A method for determining the age, rejuvenation state, or reprogramming state of cells in a sample, The steps include obtaining data on the amount of the transcript present in the aforementioned sample, A step of determining the chromatin state of the sample using the method described in any one of claims 1 to 26, The steps include comparing the biomarker value of the sample with one or more corresponding control values, Includes, Optionally, the control value is obtained from one or more samples having a known age, rejuvenation state, or reprogramming state. method.

31. A method for treating a subject diagnosed with age-related disorders, The step includes administering to the subject a therapeutically effective amount of a drug that has been identified as effective in treating the aforementioned aging-related disorders, The aforementioned drug has been identified as effective in treating the age-related disorder using the method described in any one of claims 27 to 29. method.

32. The method according to claim 31, further comprising the step of identifying the agent effective for treating the aging-related disorder using the method according to any one of claims 27 to 29, wherein the one or more test agents include the agent identified as effective for treating the aging-related disorder.

33. The method according to claim 31 or 32, wherein the control value includes the value of the biomarker metric obtained using the method according to any one of claims 1 to 26 for at least one healthy sample and at least one sample containing cells having aging-related disorders.

34. The method according to any one of claims 31 to 33, wherein the aging-related disorder is HGPS or WRN.

35. A system comprising a processor and a computer-readable medium containing instructions that, when executed by the processor, cause the processor to perform steps of any method described herein, such as the method described in any one of claims 1 to 26.

36. One or more non-temporary computer-readable media, which, when executed by one or more processors, include instructions causing the one or more processors to perform steps of any method described herein, such as the method described in any one of claims 1 to 26.

Images