Embryo development potential deep learning prediction method based on blastocyst transcription sequencing

Abstract

The invention relates to an embryonic development potential deep learning prediction method based on blastocyst transcription sequencing, which can overcome the difficulty of DNA / RNA co-sequencing for blastocyst trophoderm trace cells, and additionally obtain transcriptome RNA-seq data on the basis of not influencing the existing PGT-A technology to obtain chromosome CNV through DNA sequencing. Differential expression genes (DEGs) are found, and the developmental potential of an embryo is predicted by analyzing the expression profiles of the genes. A blastocyst pregnancy outcome prediction model is established and serves as a method for screening embryonic development potential clinically, pregnancy development potential prediction is rapidly carried out, pregnancy time can be shortened, multiple pregnancy failures are reduced, clinical efficiency is improved, and important contribution is made to the field of reproductive medicine.

Description

Technical Field

[0001] The present disclosure relates to the field of intelligent medical technologies, and particularly to a deep learning prediction method for embryo development potential based on blastocyst transcriptional sequencing, an embryo development potential prediction system, and an electronic device. Background Art

[0002] Assisted reproductive technology is an effective means to solve infertility. The three necessary conditions for successful embryo implantation are: embryos with developmental potential, an endometrium in a receptive state, and synchronous development of the embryo and the endometrium. Therefore, accurately screening the embryos with the greatest developmental potential is the key to improving the success rate of assisted reproductive technology. At present, the screening of embryo development potential widely used clinically is mainly based on traditional morphological evaluation, combined with some embryo development kinetic parameters, as well as related analysis methods such as metabolomics and genetics. Conventional pre-implantation genetic testing technology (Pre-implantation Genetic Testing for Aneuploidy, PGT-A) is a method for screening embryo development potential through DNA sequencing. Based on a comprehensive screening of copy number variations (CNVs) such as chromosomal aneuploidy, microdeletions, and microduplications of 23 pairs of chromosomes, euploid embryos suitable for implantation can be screened out, which can increase the pregnancy rate and reduce the risk of miscarriage caused by embryo chromosomal abnormalities.

[0003] A method for screening embryo development potential based on DNA / RNA co-sequencing of this product detects not only the chromosomal CNVs of embryos but also the gene expression levels of the embryo transcriptome, and uses this as a new prediction index for embryo development potential.

[0004] However, the existing PGT-A technology only detects the chromosomal CNVs of embryos at the DNA level. The true developmental potential of embryos is very complex and requires further research in multiple omics in different dimensions (morphological indicators, genetic indicators, biochemical indicators, molecular biological indicators, etc.). There have been studies on the transcriptome of early pre-implantation embryos, constructing a gene expression atlas of human embryo development, discovering a batch of new marker genes related to cell differentiation, and finding that the expression of multiple genes is abnormal in the trophectoderm cells of blastocysts with implantation failure. However, there is currently no analysis of differentially expressed genes in the transcriptome of chromosomally euploid blastocysts with different pregnancy outcomes and establishment of a prediction model for pregnancy outcomes for clinical application. Previous technologies usually only detect at the level of embryo chromosomal euploidy. However, for blastocysts with the same euploidy, their pregnancy outcomes are different, and it is impossible to accurately predict whether a blastocyst can fully develop into an individual. Therefore, it is necessary to improve the prediction means and enhance the accuracy of blastocyst development prediction. Summary of the Invention

[0005] To solve the above problems, the present application proposes a deep learning prediction method for embryo developmental potential based on blastocyst transcriptional sequencing, an embryo developmental potential prediction system, and an electronic device.

[0006] On the one hand, the present application proposes a deep learning prediction method for embryo developmental potential based on blastocyst transcriptional sequencing, including the following steps:

[0007] S1. Obtain the transcriptome data after the transfer of euploid blastocysts of patients after PGT-A treatment;

[0008] S2. Perform feature engineering on the transcriptome data to obtain corresponding differentially expressed gene features, and import them into a pre-deployed pregnancy outcome prediction model;

[0009] S3. Through the pregnancy outcome prediction model, identify and output the developmental potential prediction results of clinical pregnancy.

[0010] As an alternative implementation of the present application, optionally, the method for generating the pregnancy outcome prediction model includes:

[0011] Collect the transcriptome data after the transfer of euploid blastocysts of several patients after PGT-A treatment;

[0012] Perform gene function analysis on the transcriptome data to obtain the transcriptome expression genes of euploid blastocysts with successful clinical pregnancy and failed clinical pregnancy respectively;

[0013] Respectively screen the transcriptome differentially expressed gene features of euploid blastocysts with successful clinical pregnancy and failed clinical pregnancy, and construct corresponding differentially expressed gene feature sets A;

[0014] According to a preset ratio, divide the differentially expressed gene feature set A into: a training set A1 and a validation set A2;

[0015] Import the training set A1 into a preset machine learning model, and perform training and learning of differentially expressed gene features according to the preset model optimization and iteration conditions to obtain the initial pregnancy outcome prediction model;

[0016] Use the validation set A2 to verify the pregnancy outcome prediction model, and judge whether the pregnancy outcome prediction model is qualified:

[0017] If it is qualified, deploy the pregnancy outcome prediction model to the background server;

[0018] Conversely, retrain according to the above steps.

[0019] As an alternative implementation of the present application, optionally, the machine learning model is:

[0020] RF: Random Forest model;

[0021] SVM: Support Vector Machine;

[0022] or

[0023] LDA: Linear Discriminant Analysis model.

[0024] As an alternative implementation of this application, optionally, the predicted results of developmental potential include a good group, a medium group, and a poor group divided according to the implantation rate, where:

[0025] L (good group) ≥ 0.6,

[0026] 0.6 > L (medium group) ≥ 0.5,

[0027] 0.5 > L (poor group),

[0028] L represents the implantation rate.

[0029] On the other hand, this application proposes an embryo developmental potential prediction system, which is implemented based on the deep learning prediction method for embryo developmental potential based on blastocyst transcriptome sequencing, including:

[0030] A data acquisition unit, configured to obtain transcriptome data after the transfer of euploid blastocysts in patients who have received PGT-A treatment;

[0031] A data processing unit, configured to perform feature engineering on the transcriptome data, obtain corresponding differentially expressed gene features, and import them into a pre-deployed pregnancy outcome prediction model;

[0032] A pregnancy prediction unit, configured to identify and output the predicted results of developmental potential for clinical pregnancy through the pregnancy outcome prediction model;

[0033] The data acquisition unit is communicatively connected to the data processing unit;

[0034] The data processing unit is communicatively connected to the pregnancy prediction unit.

[0035] On the other hand, this application also proposes an electronic device, including:

[0036] A processor;

[0037] A memory for storing instructions executable by the processor;

[0038] Wherein, when the processor is configured to execute the executable instructions, it implements the deep learning prediction method for embryo developmental potential based on blastocyst transcriptome sequencing.

[0039] The technical effects of the present invention:

[0040] This application can overcome the problem of DNA / RNA co-sequencing of trace cells of blastocyst trophectoderm, and obtain additional transcriptome RNA-seq data without affecting the existing PGT-A technology to obtain chromosome CNV through DNA sequencing;

[0041] In addition to chromosome CNV detection, blastocyst transcriptome levels (transcriptome RNA-seq expression levels) can be used as a predictor of embryonic developmental potential; transcriptome levels usually refer to comprehensive analysis of RNA in embryonic cells through RNA sequencing (RNA-Seq) technology. RNA-Seq is a high-throughput sequencing technology that can accurately measure gene expression levels, discover differentially expressed genes (DEGs), and predict the developmental potential of embryos by analyzing the expression profiles of these genes.

[0042] Establishing a blastocyst pregnancy outcome prediction model as a method for clinically screening embryo development potential and quickly predicting pregnancy development potential can speed up pregnancy time, reduce multiple pregnancy failures, improve clinical efficiency, and make an important contribution to the field of reproductive medicine.

[0043] Further features and aspects of the present disclosure will become apparent from the following detailed description of exemplary embodiments with reference to the attached drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] The accompanying drawings, which are incorporated in and constitute a part of the specification, illustrate exemplary embodiments, features, and aspects of the disclosure and, together with the description, serve to explain the principles of the disclosure.

[0045] Figure 1 It is shown as a schematic diagram of the implementation process of the present invention;

[0046] Figure 2 Shown is a detection experiment diagram of differential genes between successful clinical pregnancy and unpregnant blastocysts of the present invention;

[0047] Figure 3 Shown is a diagram of the detection experiment of the transcriptome characteristics of the blastocysts on the 5th and 6th days of the present invention;

[0048] Figure 4 Shown is a schematic diagram of the receiver operating characteristic curve of the test set and validation set model of the present invention;

[0049] Figure 5 Shown is a schematic diagram of the system composition of the present invention;

[0050] Figure 6 It is a schematic diagram showing the application of the electronic device of the present invention. DETAILED DESCRIPTION

[0051] Various exemplary embodiments, features, and aspects of the present disclosure will be described in detail below with reference to the accompanying drawings. Like reference numerals in the drawings denote functionally identical or similar elements. Although various aspects of the embodiments are shown in the drawings, the drawings are not necessarily drawn to scale unless otherwise specified.

[0052] As used herein, the term "exemplary" means "serving as an example, embodiment, or illustration." Any embodiment described herein as "exemplary" is not necessarily to be construed as superior to or better than other embodiments.

[0053] In addition, for a better illustration of the present disclosure, numerous specific details are given in the following detailed description. Those skilled in the art should understand that the present disclosure can be implemented without some of these specific details. In some instances, well-known means, elements, and circuits have not been described in detail so as to highlight the gist of the present disclosure.

[0054] Embodiment 1

[0055] As Figure 1 shown, on the one hand, the present application provides a deep learning prediction method for embryo developmental potential based on blastocyst transcriptome sequencing, including the following steps:

[0056] S1. Obtain the transcriptome data after the transfer of euploid blastocysts of patients after PGT-A treatment;

[0057] S2. Perform feature engineering on the transcriptome data to obtain corresponding differentially expressed gene features, and import them into a pre-deployed pregnancy outcome prediction model;

[0058] S3. Through the pregnancy outcome prediction model, identify and output the developmental potential prediction result of clinical pregnancy.

[0059] The main purpose of the present invention is to predict whether a blastocyst can develop into an individual. Through the analysis of differentially expressed genes in the blastocyst transcriptome and model construction, the present invention obtains a pregnancy outcome prediction model for predicting and outputting the developmental potential prediction result of clinical pregnancy, which can be quickly used for clinical pregnancy prediction and has a good predictive effect on the final pregnancy outcome.

[0060] The present invention provides a screening method for embryo developmental potential based on DNA / RNA co-sequencing, mainly achieving:

[0061] Establish a DNA / RNA co-sequencing technology for a small amount of trophectoderm cells;

[0062] Rank the embryo developmental potential through the analysis of the blastocyst transcriptome level, and preferentially transplant the blastocysts with high developmental potential to improve the embryo implantation rate;

[0063] Analyze the molecular mechanism of embryo implantation failure through the transcriptome level of blastocysts with clinical pregnancy failure.

[0064] Principle of the invention:

[0065] For euploid blastocysts obtained after PGT-A treatment, analyze the transcriptome differential expression gene characteristics of euploid blastocysts with successful clinical pregnancy and failed clinical pregnancy after blastocyst transfer, and construct a pregnancy outcome prediction model based on this.

[0066] The model training and application of the present invention will be described in detail below.

[0067] The reagents or equipment involved in blastocyst biopsy, sequencing, transplantation, etc. of the present invention can all be purchased on the market or prepared in the laboratory, so they will not be limited and described in this embodiment.

[0068] The present invention adopts a retrospective analysis method, constructs a model through differentially expressed genes (DEGs), and evaluates the clinical pregnancy outcomes of euploid blastocysts with different pregnancy results.

[0069] Sample size: The study involved a total of 102 patients who underwent 111 PGT-A (preimplantation genetic testing) cycles during the period from March 2022 to July 2023, and a total of 412 blastocysts were processed.

[0070] Experimental materials: All blastocysts were biopsied on the 5th or 6th day, and a part of the sample was used for DNA copy number variation detection, and the other part was used for RNA sequencing.

[0071] Experimental steps: Blastocyst culture, selection and biopsy: Blastocyst culture was carried out according to the standard procedure, and biopsy was performed at the appropriate time point.

[0072] 1. D&R-seq: The biopsied TE cells were immediately lysed, and simultaneous sequencing of DNA and RNA was performed by D&R-seq technology.

[0073] 2. Data analysis method: Data analysis includes preprocessing of RNA-seq data, differential expression gene analysis (using the DEseq2 package), and inference of CNV (copy number variation) and protein-protein interaction network (PPI N) analysis based on RNA-seq data.

[0074] 3. Trophectoderm biopsy of blastocysts

[0075] The biopsy cells were immediately added to 7 μL of lysis buffer containing a final concentration of 1 U of RNase inhibitor for lysis. After lysis, 3.5 μL was transferred to a new PCR tube for DNA synthesis, and whole-genome amplification was continued according to the instructions of the ChromInst library preparation kit. The amplified genomes of each sample were sequenced at 0.04-fold genomic depth using the Illumina Next-seq 550 platform.

[0076] The remaining samples were subjected to RNA-seq analysis. Using a microRNA amplification kit, the mRNA of the lysed cells was reverse transcribed and amplified to synthesize the first strand of cDNA with poly(dT). Then, a linker sequence was added to the 3' end of the cDNA to complete Template-switching, and exponential amplification enrichment was performed using this linker region as the primer anchor site to obtain full-length high-quality cDNA. Finally, after enzymatic digestion and library construction, an RNA-seq library of 200 - 500 bp was obtained for sequencing, with a single-end sequencing read length of 55 bp and a data volume of 2.5 M Reads.

[0077] 4. The biopsied blastocysts were cryopreserved. According to the results of chromosomal CNV detection, euploid blastocysts were selected for thawing and transplantation.

[0078] 5. Analyze the transcriptome levels of blastocysts at different developmental times on the 5th and 6th days, search for upregulated genes (Up) and downregulated genes (Down), and perform gene function analysis (the Not group was used as the reference group). Eighty-three euploid blastocysts with pregnancy outcomes were included, and the transcriptome characteristics of euploid blastocysts with successful clinical pregnancy and failed clinical pregnancy were analyzed to screen 280 differentially expressed genes.

[0079] As shown in the Figure 2 appendix, these are the differential genes between successfully clinically pregnant and non-pregnant blastocysts. (A) Volcano plot of differential genes between the two groups. Red dots indicate upregulated genes, and blue dots indicate downregulated genes. (B, C) Gene Ontology (GO) and KEGG analyses of highly enriched molecular functions, cellular components, biological processes, and pathways between successfully clinically pregnant and non-pregnant blastocysts (p < 0.05), which are involved in the PI3K-Akt, Rap1 signaling pathways, and inflammatory mediator regulators of TRP channels.

[0080] As Figure 3As shown, for euploid blastocysts obtained after PGT-A treatment, the transcriptomic differentially expressed gene characteristics of euploid blastocysts with successful clinical pregnancy and failed clinical pregnancy after blastocyst transfer were analyzed to construct a pregnancy outcome prediction model. Among them, the transcriptomic characteristics of day 5 and day 6 blastocysts. (A) Volcano plot of DEGs between blastocysts formed on day 5 and day 6. Red dots represent genes upregulated in day 5 and day 4 blastocysts, and blue dots represent downregulated genes. (B) KEGG analysis of highly enriched pathways (p<0.05) between blastocysts formed on day 5 and day 6. (C) The top 3 significant clusters with a score > 3.0 identified using the Molecular Complex Detection (MCODE) method. Green genes are from the DEGs of day 5 and day 6 when forming blastocysts, and pink genes are key genes in the endometrial receptive phase. Genes involved in the dialogue were enriched after GO and KEGG analysis (p<0.05), and the annotations are shown on the right side of the clustering map. An euploid optimization model was constructed using the transcriptomic differences between euploid blastocysts with successful and failed clinical pregnancies. The present invention also analyzed the transcriptomic characteristics of blastocysts with different formation times (day 5 and day 6) and detected the interaction network between key blastocyst genes and the receptive endometrium.

[0081] Through the above-mentioned transcript data collection and characteristic gene analysis, the transcriptomic differentially expressed gene characteristics of euploid blastocysts with successful clinical pregnancy and failed clinical pregnancy can be obtained, which can be used for further model training next.

[0082] As an alternative implementation of the present application, optionally, the method for generating the pregnancy outcome prediction model includes:

[0083] Collecting transcriptomic data of euploid blastocysts after transplantation in several patients who have received PGT-A treatment;

[0084] Performing gene function analysis on the transcriptomic data to obtain the transcriptomic expressed genes of euploid blastocysts with successful clinical pregnancy and failed clinical pregnancy respectively;

[0085] Respectively screening the transcriptomic differentially expressed gene characteristics of euploid blastocysts with successful clinical pregnancy and failed clinical pregnancy, and constructing a corresponding differentially expressed gene characteristic set A;

[0086] Dividing the differentially expressed gene characteristic set A into a training set A1 and a validation set A2 according to a preset ratio;

[0087] Importing the training set A1 into a preset machine learning model, and performing training and learning of differentially expressed gene characteristics according to the preset model optimization and iteration conditions to obtain the initial pregnancy outcome prediction model;

[0088] Using the validation set A2 to verify the pregnancy outcome prediction model and determining whether the pregnancy outcome prediction model is qualified:

[0089] If qualified, deploy the pregnancy outcome prediction model to the background server;

[0090] On the contrary, retrain according to the above steps.

[0091] As an alternative implementation of this application, optionally, the machine learning model is:

[0092] RF: Random Forest model;

[0093] SVM: Support Vector Machine;

[0094] Or

[0095] LDA: Linear Discriminant Analysis model.

[0096] As an alternative implementation of this application, optionally, the developmental potential prediction results include a good group, a medium group, and a poor group divided according to the implantation rate, where:

[0097] L (good group) ≥ 0.6,

[0098] 0.6 > L (medium group) ≥ 0.5,

[0099] 0.5 > L (poor group),

[0100] L represents the implantation rate.

[0101] In this embodiment, 83 euploid blastocysts with pregnancy outcomes will be included and divided according to the following ratio: 59 euploid transferred blastocysts as the training set and 24 euploid transferred blastocysts as the test set.

[0102] Modeling is performed using Random Forest (RF), Support Vector Machine (SVM), and Linear Discriminant Analysis (LDA).

[0103] The feature recognition and application functions of each model will be described separately below:

[0104] 1). Random Forest (RF)

[0105] Feature type: Random Forest can handle a large number of features and is often used to process high-dimensional data. It can automatically perform feature selection and identify the most important features for prediction.

[0106] Feature extraction:

[0107] Morphological features: such as the number of cells, cell symmetry, amount of cell debris, etc.

[0108] Genetic features: such as euploidy status, genomic CNV, etc.

[0109] Biochemical characteristics: such as metabolite concentrations in embryo culture media.

[0110] Molecular biology characteristics: such as gene expression levels, differentially expressed genes (DEGs), etc.

[0111] Application of AI algorithms: Generally, the RF model itself does not require a complex feature extraction process because its internal mechanism will automatically select important features. However, in the preprocessing stage, some data dimensionality reduction techniques (such as PCA) may be used.

[0112] 2). Support Vector Machine (SVM)

[0113] Feature type: SVM is suitable for high-dimensional data, but the number of features cannot be too large, otherwise it will lead to overfitting.

[0114] Feature extraction:

[0115] Morphological characteristics: same as RF.

[0116] Genetic characteristics: same as RF.

[0117] Biochemical characteristics: same as RF.

[0118] Molecular biology characteristics: same as RF.

[0119] Application of AI algorithms: Before using SVM for modeling, feature extraction and selection are very important. It may be necessary to first perform feature dimensionality reduction or feature selection techniques (such as L1 regularization) to reduce the feature dimension.

[0120] 3). Linear Discriminant Analysis (LDA)

[0121] Feature type: LDA is suitable for a small number of features, and the features need to satisfy the normal distribution assumption.

[0122] Feature extraction:

[0123] Morphological characteristics: same as RF.

[0124] Genetic characteristics: same as RF.

[0125] Biochemical characteristics: same as RF.

[0126] Molecular biology characteristics: same as RF.

[0127] Application of AI algorithms: LDA requires more rigorous processing in feature selection. Usually, feature screening will be performed first to ensure that the features satisfy the assumptions of LDA.

[0128] For the general model training steps (taking the random forest as an example in this embodiment), the steps to construct a random forest model usually include the following stages:

[0129] 1). Data Preparation: Collect and organize the required dataset, including feature variables and target variables. The data needs to be preprocessed, such as handling missing values, outliers, as well as performing feature encoding and standardization.

[0130] 2). Feature and Sample Selection: Random Forest constructs multiple decision trees by randomly selecting features and samples from the original dataset. Usually, each tree randomly draws a certain number of samples from the original dataset and randomly selects a part of the features from all features.

[0131] 3). Decision Tree Construction: Use the decision tree algorithm to construct decision trees for each sub - dataset. During the construction process, the split of each node is based on randomly selected features and the best split criterion (such as information gain or Gini impurity).

[0132] 4). Ensemble of Decision Trees: Repeat the above process of constructing decision trees until the preset number of trees is reached. These decision trees form the Random Forest.

[0133] 5). Model Training: Use the training dataset to train the Random Forest model. During the training process, each decision tree independently learns the patterns of the dataset.

[0134] 6). Model Evaluation: Use the validation set or cross - validation method to evaluate the performance of the model. Evaluation metrics may include accuracy, recall, F1 - score, etc.

[0135] 7). Adjust Model Parameters: According to the results of model evaluation, it may be necessary to adjust model parameters, such as the number of trees, the depth of the trees, the number of features considered during splitting, etc., to optimize the model performance.

[0136] 8). Prediction and Application: Use the trained Random Forest model to make predictions on new data and apply the model to practical problems.

[0137] In this embodiment, it is verified that the AUCs of the three machine learning models are 0.88, 0.71, and 0.84 respectively.

[0138] Through the above RF model training, in this embodiment, according to the probabilities predicted by RF, the developmental potentials of 83 euploid blastocysts are divided into three groups: good, medium, and poor, and a retrospective analysis of pregnancy outcomes is carried out. The results show that the implantation rate of the group with a predicted good developmental potential is significantly higher than that of the medium group, and the implantation rate of the medium group is significantly higher than that of the poor group.

[0139] Such as Figure 4As shown, the receiver operating characteristic curves of the test set and validation set models. (A) AUC of the test set and (B) AUC of the validation set. The red line is the performance of the random forest, the blue line is the performance of the support vector machine, and the green line is the performance of LDA. (C) Retrospective analysis of the implantation rate based on the predicted probability. Good indicates that the probability of a normal embryo is equal to or greater than 0.6, medium indicates that the probability of a normal embryo is equal to 0.5 or greater than 0.5 and less than 0.6, poor indicates that the probability of a normal embryo is less than 0.5, and original indicates the true clinical implantation result. (D) Distribution of good, medium, and poor among all diploids.

[0140] Therefore, the present invention found 187 differentially expressed genes (DEGs) mainly involved in embryonic placental development, PPAR signaling pathway, and Notch signaling pathway between euploid blastocysts formed on day 5 and blastocysts formed on day 6. These DEGs have functional conversations with endometrial specific genes, indicating further differentiation of the embryo during post-implantation development.

[0141] Through retrospective analysis, it was found that machine learning models (random forest, support vector machine, and linear discriminant analysis) constructed based on 280 differentially expressed genes could effectively optimize the selection of euploid blastocysts, and the AUC of the RF model reached 0.88.

[0142] Quantitative results: Experimental data showed that the implantation rate of blastocysts in the high-quality group was significantly higher than that in the medium-quality group (88.6% vs. 50%, p = 0.001), and the medium-quality group was also significantly higher than that in the low-quality group (50% vs. 20.8%, p = 0.035).

[0143] Therefore, it was confirmed that in addition to chromosomal euploidy, the transcriptome level of blastocysts is also an important indicator for predicting embryo implantation potential.

[0144] The present invention developed a new D&R-seq technology that can simultaneously evaluate the genome and transcriptome of blastocysts, providing a multi-dimensional reference for optimizing embryo selection. This optimization system is expected to shorten the pregnancy time, reduce multiple pregnancy failures, improve clinical efficiency, and make important contributions to the field of reproductive medicine.

[0145] Obviously, those skilled in the art should understand that to implement all or part of the processes in the above embodiments, it can be completed by instructing relevant hardware through a computer program. The program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the above control embodiments. Those skilled in the art can understand that to implement all or part of the processes in the above embodiments, it can be completed by instructing relevant hardware through a computer program. The program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the above control embodiments. Among them, the storage medium can be a magnetic disk, an optical disc, a read-only memory (ROM), a random access memory (RAM), a flash memory, a hard disk drive (HDD), or a solid-state drive (SSD), etc.; the storage medium can also include a combination of the above types of memories.

[0146] Embodiment 2

[0147] As Figure 5 shown, based on the implementation principle of Embodiment 1, on the other hand, this application proposes an embryo developmental potential prediction system, which is implemented based on the deep learning prediction method for embryo developmental potential based on blastocyst transcriptome sequencing, and includes:

[0148] A data acquisition unit, configured to obtain transcriptome data after the transfer of euploid blastocysts in patients after receiving PGT-A treatment;

[0149] A data processing unit, configured to perform feature engineering on the transcriptome data, obtain corresponding differentially expressed gene features, and import them into a pre-deployed pregnancy outcome prediction model;

[0150] A pregnancy prediction unit, configured to identify and output a developmental potential prediction result of clinical pregnancy through the pregnancy outcome prediction model;

[0151] The data acquisition unit is communicatively connected to the data processing unit;

[0152] The data processing unit is communicatively connected to the pregnancy prediction unit.

[0153] For the functions and interactions of the above-mentioned various units, please refer to Embodiment 1 for understanding.

[0154] Each module or step of the present invention described above can be implemented by a general-purpose computing system. They can be concentrated on a single computing system or distributed over a network composed of multiple computing systems. Optionally, they can be implemented by program code executable by the computing system. Thus, they can be stored in the storage system and executed by the computing system, or they can be separately fabricated into individual integrated circuit modules, or multiple modules or steps among them can be fabricated into a single integrated circuit module for implementation. In this way, the present invention is not limited to any specific combination of hardware and software.

[0155] Embodiment 3

[0156] As Figure 6 shown, further, on the other hand, the present application also proposes an electronic device, including:

[0157] A processor;

[0158] A memory for storing instructions executable by the processor;

[0159] Wherein, when the processor is configured to execute the executable instructions, it implements a deep learning prediction method for embryo development potential based on blastocyst transcriptome sequencing described in Embodiment 2.

[0160] The electronic device of the embodiments of the present disclosure includes a processor and a memory for storing instructions executable by the processor. Wherein, when the processor is configured to execute the executable instructions, it implements a deep learning prediction method for embryo development potential based on blastocyst transcriptome sequencing described in Embodiment 2 above.

[0161] Here, it should be noted that the number of processors can be one or more. At the same time, in the electronic device of the embodiments of the present disclosure, an input system and an output system can also be included. Among them, the processor, the memory, the input system and the output system can be connected through a bus or in other ways, which is not specifically limited here.

[0162] As a computer-readable storage medium, the memory can be used to store software programs, computer-executable programs and various modules, such as: programs or modules corresponding to a deep learning prediction method for embryo development potential based on blastocyst transcriptome sequencing of the embodiments of the present disclosure. The processor executes various functional applications and data processing of the electronic device by running the software programs or modules stored in the memory.

[0163] The input system can be used to receive input numbers or signals. Among them, the signal can be a key signal related to the user settings and function control of the device / terminal / server. The output system can include a display device such as a display screen.

[0164] The embodiments of the present disclosure have been described above. The above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and variations will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The choice of terms used herein is intended to best explain the principles of the embodiments, practical applications, or improvements to technologies in the market, or to enable other ordinary skill in the art to understand the embodiments disclosed herein.

Claims

1. A deep learning prediction method for embryonic developmental potential based on blastocyst transcriptome sequencing, characterized in that It includes the following steps: S1. Obtain the transcriptome data after the transfer of euploid blastocysts in patients who have received PGT-A treatment; S2. Perform feature engineering on the transcriptome data to obtain the corresponding differentially expressed gene features and import them into a pre-deployed pregnancy outcome prediction model; S3. Through the pregnancy outcome prediction model, identify and output the prediction results of the developmental potential of clinical pregnancy.

2. The method for predicting the embryonic development potential by deep learning based on blastocyst transcriptome sequencing according to claim 1, wherein The method for generating the pregnancy outcome prediction model includes: Collect the transcriptome data after the transfer of euploid blastocysts in several patients who have received PGT-A treatment; Perform gene function analysis on the transcriptome data to obtain the transcriptome expression genes of euploid blastocysts with successful clinical pregnancy and failed clinical pregnancy respectively; Screen the differentially expressed gene features of the transcriptomes of euploid blastocysts with successful clinical pregnancy and failed clinical pregnancy respectively, and construct the corresponding differentially expressed gene feature set A; According to a preset ratio, divide the differentially expressed gene feature set A into: a training set A1 and a validation set A2; Import the training set A1 into a preset machine learning model, and perform training and learning of the differentially expressed gene features according to the preset model optimization and iteration conditions to obtain the initial pregnancy outcome prediction model; Use the validation set A2 to verify the pregnancy outcome prediction model and determine whether the pregnancy outcome prediction model is qualified: If it is qualified, deploy the pregnancy outcome prediction model to the background server; Otherwise, retrain according to the above steps.

3. The method for predicting the embryonic developmental potential by deep learning based on blastocyst transcriptome sequencing according to claim 2, wherein The machine learning model is: RF: Random Forest model; SVM: Support Vector Machine; Or LDA: Linear Discriminant Analysis model.

4. The method for predicting the embryonic development potential by deep learning based on blastocyst transcriptome sequencing according to claim 1, wherein The prediction results of the developmental potential include a good group, a medium group, and a low group divided according to the implantation rate, where: L (good group) ≥ 0.6, 0.6 > L (medium group) ≥ 0.5, 0.5 > L (low group), L represents the implantation rate.

5. An embryonic development potential prediction system, which is implemented based on the deep learning prediction method for embryonic development potential based on blastocyst transcriptional sequencing according to any one of claims 1-4, characterized in that, It includes: A data acquisition unit for obtaining the transcriptome data after the transfer of euploid blastocysts in patients who have received PGT-A treatment; A data processing unit for performing feature engineering on the transcriptome data to obtain the corresponding differentially expressed gene features and importing them into a pre-deployed pregnancy outcome prediction model; A pregnancy prediction unit for identifying and outputting the prediction results of the developmental potential of clinical pregnancy through the pregnancy outcome prediction model; The data acquisition unit is communicatively connected to the data processing unit; The data processing unit is communicatively connected to the pregnancy prediction unit.

6. An electronic device, characterized in that, It includes: A processor; A memory for storing instructions executable by the processor; Wherein, when the processor is configured to execute the executable instructions, it implements the deep learning prediction method for embryo developmental potential based on blastocyst transcriptome sequencing according to any one of claims 1-4.

Images