Methods, apparatus, equipment, media and products for identifying target interventions

Abstract

This application provides a method, apparatus, device, medium, and product for determining target interventions, relating to the field of bioinformatics. The method includes: acquiring N sets of gene expression feature sets corresponding to K interventions, and M whole-genome gene expression feature profiles corresponding to a predetermined phenotype in a reference database; for the k-th intervention, performing mathematical transformations on its N sets of gene expression feature sets and the M whole-genome gene expression feature profiles respectively to obtain reverse matching results; extracting representative values ​​from the reverse matching results of the same intervention under the same phenotypic feature profile to form M target results; aggregating the M target results of the k-th intervention to obtain representative results for that intervention; and sorting the representative results of the K interventions to determine the target intervention. This application's solution can significantly improve the accuracy of intervention prediction and reduce the risk of missed screening.

Description

Technical Field

[0001] This application relates to the field of bioinformatics, and in particular to methods, devices, electronic devices, computer-readable storage media, and computer program products for identifying target interventions. Background Technology

[0002] With the widespread adoption of high-throughput transcriptome sequencing technology and the rapid accumulation of public transcriptome data, the identification of candidate interventions based on gene expression profiles has become an important strategy for accelerating the discovery of new interventions and the repurposing of existing ones. This technology compares the target phenotypic state with gene expression changes after intervention perturbation to achieve systematic prediction and prioritization of potential candidate interventions, providing an efficient and low-cost computational screening approach for drug development, agricultural conservation, toxicity assessment, and other fields.

[0003] Among related technologies, intervention screening methods based on gene expression mainly rely on Connectivity Map (CMap) and its derived L1000 platform. This technology establishes a reference database of gene expression characteristics of small molecule compounds under different cell lines, concentrations, and treatment times. Users provide disease-related upregulated and downregulated gene sets, and screening is achieved by calculating connectivity scores or reversal scores.

[0004] However, the relevant technologies remain stuck in a "phenotype-based intervention search" model. This means that only phenotype-related upregulated / downregulated gene sets can be used as query input, and matching and screening are performed in a reference database of intervention gene expression characteristics. This query paradigm makes the prediction results highly dependent on the completeness and representativeness of the selected phenotype gene set. Once the input gene set comes from a single study, a single tissue source, or a specific phenotype stage, it cannot effectively integrate heterogeneous data of the same phenotype from different sources, platforms, or conditions. This results in unstable prediction results and easily misses truly effective candidate interventions. Summary of the Invention

[0005] The embodiments of this application aim to at least partially solve one of the aforementioned technical problems or at least provide a useful commercial option. In view of this, the embodiments of this application provide a method, apparatus, device, medium, and product for determining target interventions. Through a reverse lookup paradigm of "intervention-based phenotypic lookup," the acquisition and aggregation of whole-genome multi-source gene expression feature profiles, and the extraction of representative values ​​from the reverse matching results under each feature profile, a comprehensive assessment of the reversal potential of candidate interventions is achieved, effectively improving the accuracy of intervention prediction and reducing the risk of missed screening.

[0006] In a first aspect, embodiments of this application provide a method for determining a target intervention, the method comprising: acquiring N sets of gene expression feature sets for K interventions, and acquiring M whole-genome gene expression feature profiles corresponding to a predetermined phenotype from a reference database, wherein K, N, and M are all integers not less than 1; for the k-th intervention, based on mathematical transformation, determining the reverse matching results of the N sets of gene expression feature sets relative to the m-th whole-genome gene expression feature profile to obtain a target result, wherein the target result is a representative value of the reverse matching result, k is an integer not less than 1 and not greater than K, and m is an integer not less than 1 and not greater than M; aggregating the M target results of the k-th intervention to obtain a representative result of the k-th intervention, wherein the M target results are obtained by traversing the M whole-genome gene expression feature profiles; sorting the representative results of the K interventions to determine the target intervention, wherein the representative results of the K interventions are obtained by traversing the K interventions.

[0007] In an exemplary embodiment, the target result is a negative value.

[0008] In an exemplary embodiment, the step of determining the reverse matching results of N gene expression feature sets relative to the m-th whole-genome gene expression feature spectrum based on mathematical transformation for the k-th intervention to obtain the target result includes: dividing the n-th gene expression feature set to obtain the n-th upregulated gene set and the n-th downregulated gene set, where n is an integer not less than 1 and not greater than N; determining the enrichment scores of the n-th upregulated gene set and the n-th downregulated gene set respectively; determining the reverse matching results of the n-th gene expression feature set relative to the m-th whole-genome gene expression feature spectrum based on the difference between the enrichment scores of the n-th upregulated gene set and the n-th downregulated gene set; traversing the N gene expression feature sets to obtain N reverse matching results; and determining the representative value among the N reverse matching results to obtain the target result.

[0009] In an exemplary embodiment, determining the representative value among the N reverse matching results to obtain the target result includes: performing mathematical calculations on the N reverse matching results to determine the representative value among the N reverse matching results to obtain the target result. The mathematical calculations include: minimum value, mean, median, or weighted average.

[0010] In an exemplary embodiment, determining the representative value among the N reverse matching results to obtain the target result includes: using the minimum value among the N reverse matching results as the representative value to obtain the target result.

[0011] In an exemplary embodiment, the method further includes: standardizing the reverse matching result to obtain a standard reverse matching result, wherein the standardization process includes Min-Max standardization.

[0012] In an exemplary embodiment, the aggregation process of the M target results of the k-th intervention to obtain the representative result of the k-th intervention includes: averaging the M target results of the k-th intervention to obtain the representative result of the k-th intervention.

[0013] In an exemplary embodiment, the reference database is constructed as follows: the raw transcriptome sequencing data is subjected to quality control processing to obtain high-quality sequencing reads; the high-quality sequencing reads are aligned to a reference genome to generate a comparison file; based on the comparison file, the expression level of each gene in each sample is estimated; differential expression analysis is performed on the expression level of each gene in each sample in the experimental group and the control group to determine the whole-genome gene expression profile, wherein the whole-genome gene expression profile includes the differential expression fold change and the correction p-value for each gene.

[0014] In an exemplary embodiment, the N sets of gene expression feature sets are upregulated gene sets and downregulated gene sets induced by the intervention under different perturbation conditions, wherein the perturbation conditions include at least one of: cell line, treatment concentration, and treatment time.

[0015] In an exemplary embodiment, the intervention includes: a compound, the compound including at least one of a drug, a pesticide, and a growth regulator; the predetermined phenotype includes: a disease.

[0016] Secondly, embodiments of this application provide a target intervention determination device, which includes: a signal acquisition module, a reverse matching module, an aggregation module, and a decision module.

[0017] The system includes the following modules: a signal acquisition module for acquiring N sets of gene expression feature sets for K interventions and M whole-genome gene expression feature profiles corresponding to a predetermined phenotype from a reference database, where K, N, and M are all integers not less than 1; a reverse matching module for determining the reverse matching results of the N sets of gene expression feature sets relative to the m-th whole-genome gene expression feature profile for the k-th intervention based on mathematical transformations, to obtain the target result, where the target result is the representative value of the reverse matching result, k is an integer not less than 1 and not greater than K, and m is an integer not less than 1 and not greater than M; an aggregation module for aggregating the M target results of the k-th intervention to obtain the representative result of the k-th intervention, where the M target results are obtained by traversing the M whole-genome gene expression feature profiles; and a decision module for sorting the representative results of the K interventions to determine the target intervention, where the representative results of the K interventions are obtained by traversing the K interventions.

[0018] Thirdly, embodiments of this application provide an electronic device, which includes: a processor and a memory; the memory is used to store a computer program; the processor is used to execute the computer program to implement the target intervention determination method as provided in the first aspect.

[0019] Fourthly, embodiments of this application provide a computer-readable storage medium storing computer instructions or programs that, when executed on a computer, cause the target intervention determination method provided in the first aspect to be performed.

[0020] Fifthly, embodiments of this application provide a computer program product including computer instructions that, when some or all of the computer instructions are executed on a computer, cause the target intervention determination method provided in the first aspect to be performed.

[0021] In a sixth aspect, embodiments of this application provide a computer program that, when run on a computer, causes the computer to perform the target intervention determination method as provided in the first aspect.

[0022] In summary, the solution provided in this application obtains M whole-genome gene expression feature profiles corresponding to a predetermined phenotype from a reference database and performs reverse matching using N sets of gene expression feature sets of the intervention as query input, thus realizing a paradigm shift from the traditional "finding intervention by phenotype" to "finding phenotype by intervention". At the same time, a mechanism of "taking the representative value of the reverse matching result under each feature profile and aggregating across M profiles" is adopted to robustly integrate multi-source heterogeneous data of the same phenotype.

[0023] On the one hand, since the query input no longer relies on a single, incomplete set of upregulated / downregulated genes provided by the user, but directly utilizes M genome-wide standardized feature profiles pre-stored in the reference database, it completely eliminates the problem of incomplete data caused by the input gene set originating from a single study, a single tissue source, or a specific phenotypic stage. This allows for comprehensive coverage of the heterogeneous expression patterns of the same phenotype under different sources, platforms, and conditions, avoiding systematic bias in prediction results caused by input selection.

[0024] On the other hand, by calculating the inverse matching result for each intervention under each feature spectrum and taking its representative value, and then aggregating the M representative values ​​under M spectra, automatic and robust fusion of multi-source heterogeneous data is achieved. This effectively mitigates the influence of noise, batch effect or insufficient representativeness of a single feature spectrum, so that the final representative result no longer changes drastically with the fluctuation of the selected gene set or a single data source. This significantly improves the stability and reliability of intervention prediction and greatly reduces the risk of truly effective candidate interventions being missed. Attached Figure Description

[0025] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0026] Figure 1 This is a flowchart illustrating a method for determining a target intervention according to this application; Figure 2 This is a schematic diagram of a target intervention determination device according to an embodiment of this application; Figure 3 This is a schematic block diagram of an electronic device according to an embodiment of this application; Figure 4 This is a schematic diagram of the comparison curve of the effective drug identification ratio within the Top N according to an embodiment of this application; Figure 5 A bar chart comparing the number of known therapeutic drugs found using different methods according to one embodiment of this application. Detailed Implementation

[0027] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0028] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in sequences other than those illustrated or described herein. In embodiments of this application, "B corresponding to A" means that B is associated with A. In one implementation, B can be determined based on A. However, it should also be understood that determining B based on A does not mean determining B solely based on A; B can also be determined based on A and / or other information. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or server that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to these processes, methods, products, or devices. In the description of this application, unless otherwise stated, "a plurality of" means two or more.

[0029] In the embodiments of this application, the terms "module" or "unit" refer to a computer program or part of a computer program that has a predetermined function and works with other related parts to achieve a predetermined goal, and can be implemented wholly or partially using software, hardware (such as processing circuitry or memory), or a combination thereof. Similarly, a processor (or multiple processors or memory) can be used to implement one or more modules or units. Furthermore, each module or unit can be part of an overall module or unit that includes the functionality of that module or unit.

[0030] Before introducing the technical solution of this application, the relevant knowledge of this application will be introduced below: In this application, "intervention" refers to any substance or entity that can induce detectable changes in gene expression in cells or organisms. In the embodiments of this application, interventions include not only compounds in the traditional sense, but also drugs (such as extracts of traditional Chinese medicine, biological agents, cell therapy products), pesticides, or growth regulators.

[0031] In this application, a "gene expression feature set" refers to the set of differentially expressed genes relative to a control sample after treatment with an intervention under specific experimental conditions. It typically includes an upregulated gene set (genes with log2FC greater than 1 compared to the control group) and a downregulated gene set (genes with log2FC less than -1 compared to the control group). In the embodiments of this application, each gene expression feature set corresponds to a perturbation condition of the intervention (such as different cell lines, concentrations, or treatment times), which is used as input for a reverse lookup.

[0032] In this application, "predicted phenotype" refers to a predicted or screened target biological or pathological state, characterized by differential expression data from the whole transcriptome. In the embodiments of this application, the predicted phenotype mainly includes human diseases, but can also be extended to toxicity phenotypes, plant disease phenotypes, etc. For example, breast cancer, liver toxicity, or rice disease resistance can all be used as predicted phenotypes, and their corresponding M gene expression profiles are stored in a reference database.

[0033] In this application, "whole-genome gene expression profile" refers to the differential expression values ​​(log2FC) of all genes across the entire genome and their global ranking list obtained through standardized transcriptome analysis for a predetermined phenotype. In the embodiments of this application, each of the M profiles covers approximately 60,000 human genes, which is different from the limitation of the existing L1000 platform that only directly measures 978 genes, ensuring the integrity and comparability of the data.

[0034] In this application, "reverse matching result" refers to the quantitative value of the degree of reversal (i.e., RGES) obtained based on enrichment statistics between the gene expression feature set of the intervention and the genome-wide gene expression feature profile of the predetermined phenotype. The more negative the value, the stronger the ability of the intervention to reverse the gene expression pattern of that phenotype. In the embodiments of this application, the reverse matching result is determined by the difference in enrichment scores between the upregulated gene set and the downregulated gene set.

[0035] In this application, "target result" refers to the representative value selected from the N reverse matching results corresponding to the N gene expression feature sets of the intervention for each m-th whole-genome gene expression feature profile. In the embodiments of this application, the target result is usually a negative value, and its function is to extract the optimal reversal signal of the intervention under that feature profile.

[0036] In this application, "representative value" refers to the most representative value determined from N reverse matching results. In this application, the representative value can be obtained through mathematical calculations such as minimum, mean, median, or weighted average.

[0037] In this application, "aggregation processing" refers to the process of statistically aggregating M target results obtained from the same intervention under M whole-genome gene expression feature profiles. In the embodiments of this application, aggregation processing includes forms such as average aggregation, used to generate the final representative result (rRGES) and achieve robust integration across multi-source heterogeneous data.

[0038] In this application, "standardization processing" refers to the process of scaling the reverse matching results. In the embodiments of this application, the Min-Max standardization method is used to map all original reverse matching results under the same phenotype to the interval [-1, 1], eliminating the numerical scale differences between different feature spectra and ensuring the fairness and comparability of subsequent aggregation processing.

[0039] In this application, "reference database" refers to a standardized database that is pre-constructed and stores a large number of genome-wide gene expression profiles corresponding to predetermined phenotypes. In the embodiments of this application, its construction process includes quality control processing, sequence alignment, expression level estimation, and differential expression analysis, ultimately generating M feature profiles covering the entire genome to support reverse lookup.

[0040] In this application, "perturbation conditions" refers to the combination of experimental parameters used when treating samples with an intervention. In the embodiments of this application, perturbation conditions include at least one of cell line, treatment concentration, and treatment time. The same intervention will generate multiple sets of gene expression feature sets under different perturbation conditions, thereby providing multiple candidate reversal signals for each feature profile.

[0041] In this application, "enrichment score" refers to a quantitative indicator of the enrichment degree of a gene set in a genome-wide sequencing list, calculated using conventional statistical methods in the art (e.g., a modified Kolmogorov-Smirnov method), including the enrichment score of upregulated gene sets (ES_up) and the enrichment score of downregulated gene sets (ES_down). In the embodiments of this application, the difference between the two is directly used to determine the reverse matching result.

[0042] The technical problems to be solved, the inventive concept and the system architecture of the embodiments of this application will be described below: Currently, intervention screening methods based on gene expression mainly rely on Connectivity Map (CMap) and its derivative L1000 platform. Besides the issues mentioned above, this platform also suffers from insufficient gene coverage. Specifically, the small molecule compound gene expression characteristic reference database established by the L1000 platform only directly detects the expression levels of 978 genes and infers the expression levels of 11,350 genes through computational methods. However, humans have more than 60,000 genes, resulting in a significant lack of gene expression data. This may lead to the loss of a large amount of key gene information related to diseases, failing to comprehensively reflect the gene change status of diseases and affecting the accuracy of predictions.

[0043] To address the aforementioned issues, this paper proposes a method that constructs and directly utilizes a whole-genome gene expression profile reference database, and uses the gene expression profile set of interventions as query input for reverse matching. This solves the problem that the existing L1000 platform, which only covers approximately 12,328 genes, results in the loss of a large amount of key gene information and cannot fully reflect the state of gene changes in diseases.

[0044] Specifically, firstly, M genome-wide gene expression profiles corresponding to the predetermined phenotype are obtained from a reference database. Each profile contains the differential expression values ​​(log2FC) of all genes (approximately 60,000) across the entire human genome, along with their global ranking list, rather than relying solely on the 978 genes directly detected and the inferred 11,350 genes detected by the L1000 platform. This avoids the problem of significant loss of key disease-related gene information due to insufficient gene coverage in related technologies, thus achieving a complete and comprehensive characterization of the gene change status of the predetermined phenotype, which helps to fundamentally improve the accuracy of intervention prediction.

[0045] Secondly, for each intervention, the reverse matching results of its N sets of gene expression feature sets relative to each whole-genome gene expression feature profile are determined based on mathematical transformations, and the representative value of this result is used as the target result. The upregulated / downregulated gene sets of the intervention can be globally enriched in the whole-genome sequencing list, rather than being limited to the finite gene space of L1000, ensuring that even if key disease genes are not in the traditional list of 978 core genes, they can still be accurately captured and participate in the matching. This avoids the "loss of key gene information" defect caused by the incomplete gene coverage of the L1000 platform, thus enabling the reverse matching results to truly reflect the ability of the intervention to reverse the whole-genome expression pattern of the disease, significantly improving the comprehensiveness and reliability of the prediction.

[0046] Next, the M target results obtained for each intervention under M whole-genome feature profiles are aggregated to obtain representative results, and the representative results of all interventions are sorted to determine the target intervention. Through aggregation operations across M whole-genome profiles, multi-source and fully covered data signals are further integrated, and even if individual genes are not reflected in a single profile due to coverage gaps, they can be effectively compensated for in multi-profile aggregation.

[0047] This effectively eliminates the prediction bias and instability issues caused by insufficient gene coverage in existing technologies, thereby achieving higher accuracy in prioritizing candidate interventions and significantly reducing the risk of missing effective interventions due to the loss of key gene information.

[0048] The technical solution of this application will be described in detail below: Figure 1 This is a flowchart of a target intervention determination method according to an embodiment of this application, with reference to... Figure 1 The method may include: S110, obtain N sets of gene expression feature sets for K interventions, and obtain M whole-genome gene expression feature profiles corresponding to the predetermined phenotype from the reference database, where K, N, and M are all integers not less than 1; This step is used to prepare the input data required for the reverse query, including multiple sets of gene expression features on the intervention side and multiple genome-wide gene expression feature profiles on the phenotype side.

[0049] In an exemplary embodiment, this step is implemented by: acquiring K interventions to be screened (K≥1), each intervention corresponding to N sets (N≥1) of gene expression feature sets. The gene expression feature set refers to the set of differentially expressed genes relative to the control group after treatment of cell or tissue samples with the intervention under specific perturbation conditions, typically divided into upregulated gene sets (log2FC>1) and downregulated gene sets (log2FC<-1). These feature sets can be derived from public databases (such as L1000FWD, LINCS, etc.) or perturbation experimental data generated by the laboratory itself.

[0050] In this embodiment, the N gene expression feature sets are derived from experimental results of the same intervention under different perturbation conditions. Perturbation conditions include, but are not limited to, at least one combination of different cell lines (e.g., MCF7, A549, HepG2, etc.), different treatment concentrations (e.g., 1 μM, 10 μM, 100 μM), and different treatment times (e.g., 3 h, 6 h, 24 h). By introducing multiple perturbation conditions, the diversity of transcriptomic responses of the intervention under different biological backgrounds can be captured, thereby providing richer candidate signals for subsequent reverse matching and helping to improve the comprehensiveness and robustness of the screening.

[0051] Simultaneously, from the reference database pre-constructed in this application, M (M≥1) whole-genome gene expression profiles corresponding to the current predetermined phenotype are obtained. Each profile corresponds to an independent phenotype dataset (usually from different patient cohorts, tissue sources, pathological stages, or sequencing platforms), containing the log2 (FC) values ​​of all genes across the whole genome and their global sorted list in descending order.

[0052] By simultaneously acquiring N feature sets from the intervention side and M whole-genome feature profiles from the phenotype side, this step establishes the basis for reverse lookup of "intervention-based phenotype lookup". This avoids the over-reliance on single, incomplete phenotype gene sets in the traditional "phenotype-based intervention-based lookup" model, thus laying a data foundation for the robust integration of subsequent multi-source heterogeneity data and helping to significantly improve the stability of prediction results and the recall rate of candidate interventions.

[0053] In an exemplary embodiment, the reference database is constructed as follows: The raw transcriptome sequencing data undergoes quality control processing to obtain high-quality sequencing reads; these high-quality reads are aligned to a reference genome to generate a comparison file; based on the comparison file, the expression level of each gene in each sample is estimated; differential expression analysis is performed on the expression levels of each gene in each sample from the experimental and control groups to determine the whole-genome gene expression profile, which includes the differential expression fold change and corrected p-value for each gene. The reference database constructed through this unified process can eliminate batch effects, sequencing platform differences, and sample processing variations between different studies, thereby providing a highly standardized and directly comparable multi-source whole-genome profile for subsequent reverse queries.

[0054] The raw transcriptome sequencing data in this embodiment can be from publicly available data, such as GEO, SRA, GDC, GTEx, and laboratory self-testing data; the above data alignment, gene expression estimation, and differential expression analysis can be calculated using conventional analysis procedures in the field, and will be described in detail below with specific examples.

[0055] For example, constructing a reference database based on the above method specifically includes: 1) Quality control processing is performed on raw RNA-Seq sequencing data (Fastq format) downloaded from public transcriptome databases (including but not limited to GEO, SRA, GDC, GTEx, etc.). First, FastQC software is used to assess the quality of the raw data, including indicators such as base quality distribution, GC content, adapter contamination, and repetitive sequences. Then, Trimmomatic software is used to remove low-quality reads (usually with a Phred quality threshold of <20) and adapter sequences to obtain high-quality sequencing reads.

[0056] 2) Align the high-quality reads mentioned above to the human reference genome (e.g., GRCh38) using STAR software to generate alignment files in BAM format. This step employs a double-pass alignment mode to improve the accuracy of splice site detection and supports multi-threaded processing for faster processing.

[0057] 3) Based on the alignment file, the expression level of each gene in each sample was estimated using RSEM software (output in TPM or FPKM / expected count format) to obtain a gene-level quantitative matrix.

[0058] 4) Differential expression analysis was performed on the experimental group (disease / phenotype group) and the control group (normal group) using DESeq2 software. A negative binomial distribution model was used to fit the count data, and the log2 (fold change) value (log2FC) and adjusted p-value of each gene were calculated. Subsequently, all genes were globally sorted in descending order based on the log2FC values ​​to generate a complete genome-wide gene expression profile, which includes the differential expression fold change and adjusted p-value information for each gene.

[0059] In this embodiment, a total of 2,652 feature profiles covering 749 human diseases were constructed. The construction method of this embodiment solves the problems of insufficient gene coverage (only directly detecting 978 genes) and inability to integrate data heterogeneity of the existing L1000 platform from the data source, providing a solid foundation for achieving robust and accurate intervention screening.

[0060] In an exemplary embodiment, the N sets of gene expression feature sets are derived from upregulated and downregulated gene sets induced by the intervention under different perturbation conditions. The perturbation conditions include at least one of cell line, treatment concentration, and treatment time.

[0061] For example, when the intervention is a small molecule compound, treatment with 10 μM for 6 hours in the MCF7 cell line yields the first set of features (upregulated gene set + downregulated gene set); subsequent treatment with 10 μM for 24 hours in the same cell line yields the second set of features; and treatment with 5 μM for 6 hours in the A549 cell line yields the third set of features. By introducing multiple perturbation conditions, the transcriptome response patterns of the intervention under different biological backgrounds can be more comprehensively characterized, thus providing multiple candidate reversal signals in the subsequent matching of each phenotypic feature profile, significantly improving the capture probability of the "strongest reversal signal" and the robustness of screening.

[0062] In an exemplary embodiment, the intervention includes compounds, specifically at least one of drugs (such as metformin, aspirin, etc.), pesticides (such as imidacloprid, carbendazim, etc.) or plant growth regulators (such as gibberellin, ethephon, etc.).

[0063] The predetermined phenotypes preferably include human diseases, such as breast cancer, Alzheimer's disease, type 2 diabetes, liver fibrosis, etc., and can also be extended to toxic phenotypes (such as hepatotoxicity, nephrotoxicity) or plant disease phenotypes (such as rice sheath blight, Arabidopsis gray mold, etc.).

[0064] For example, when the predetermined phenotype is breast cancer, multiple genome-wide feature profiles can be extracted from reference databases, including tumor tissue, blood samples, and patient-derived xenograft models. At the same time, multiple candidate drugs can be selected as interventions to generate N sets of gene expression feature sets under different cell lines and concentration conditions.

[0065] S120, for the k-th intervention, based on mathematical transformation, determine the reverse matching results of N groups of gene expression feature sets relative to the m-th whole genome gene expression feature spectrum, and obtain the target result, wherein the target result is the representative value of the reverse matching result, k is an integer not less than 1 and not greater than K, and m is an integer not less than 1 and not greater than M. This step involves quantitatively evaluating the reversal potential of the k-th intervention under different perturbation conditions (N sets of gene expression feature sets) for each predetermined phenotype's whole-genome gene expression feature profile (m-th profile), and extracting the most representative reversal signal as the "target result" for that feature profile.

[0066] In an exemplary embodiment, the target result is a negative value; that is, it is only considered a valid target result under the m-th feature spectrum if the representative value among the N reverse matching results is less than 0. If the representative value is greater than or equal to 0, the feature spectrum does not produce a valid target result or is marked as having no reversal signal. By limiting the target result to negative values, signals with clear reversal potential are strictly screened, avoiding interference from weak signals (positive values, indicating similar or aggravated phenotypes) or near-zero values ​​(no significant reversal) in subsequent aggregation steps, thereby improving the biological significance of the representative results and the specificity of the prediction. In practical applications, such as screening candidate drugs for breast cancer, compounds that merely mimic disease states or have no therapeutic potential can be effectively excluded, significantly reducing the false positive rate and making the final ranking more focused on truly effective reversal candidates.

[0067] In an exemplary embodiment, the determination of the reverse matching result for the k-th intervention and the m-th whole-genome gene expression profile includes the following sub-steps: S1. The gene expression feature set of group n is divided into the upregulated gene set and the downregulated gene set of group n. Usually, upregulated (positive value) and downregulated (negative value) genes are separated from the differentially expressed gene list based on the log2FC threshold (e.g., |log2FC|>1 and adjusted p value<0.05).

[0068] S2, determine the enrichment scores ES_up and ES_down of the nth upregulated gene set and the nth downregulated gene set in the global ranking list of the mth feature spectrum, respectively. The calculation adopts the improved Kolmogorov-Smirnov statistical method: traverse the position of upregulated / downregulated genes in the ranking list, calculate the maximum positive value a and the maximum negative value b of the cumulative bias, and take max(a, b) as the corresponding enrichment score.

[0069] For example, suppose the intervention is a small molecule compound and the predetermined phenotype is a disease; the nth set of gene expression features contains N drug The genome-wide sequence list of the m-th feature profile contains N significantly differentially expressed genes (the sum of the upregulated and downregulated gene sets). disease There are approximately 61,544 genes (covering the entire genome). Projecting all genes in the nth gene expression feature set onto a globally descending sorted list of the mth feature spectrum yields the position S of each gene in the sorted list. disease (i) (i=1,2,...,N) drug ).

[0070] The calculation process is as follows: For upregulated gene sets (up): Equation 1); Equation 2); Where, N drug N is the number of genes in a small molecule compound gene cluster. disease The number of genes in the disease signature spectrum. disease (i) represents the position of the i-th gene in the small molecule compound gene set within the disease feature spectrum ranking list. Iterate through the genes in the small molecule compound gene set, calculating a and b for each, and selecting the maximum values ​​as the final a and b. Take the larger of a and b as the enrichment score ES_up for the upregulated gene set.

[0071] For downregulating gene sets: The cumulative bias is calculated using the same formulas 1) and 2), and the maximum value of a and b is taken as the enrichment score ES_down of the downregulated gene set.

[0072] S3 determines the reverse matching result based on the difference in enrichment scores between the upregulated and downregulated gene sets. That is, RGES_n = ES_up - ES_down.

[0073] S4, iterate through N sets of gene expression feature sets to obtain N reverse matching results.

[0074] S5, determine the representative value among the N reverse matching results to obtain the target result under this feature spectrum.

[0075] The above steps realize a complete computational method from the original gene set to the reversal quantification score. This achieves precise, direction-sensitive quantification of reversal enrichment, ensuring effective capture of potential therapeutic signals even under perturbation conditions with small changes in gene expression. In application scenarios, such as evaluating multiple feature sets of a certain traditional Chinese medicine extract at different concentrations in HepG2 cells, this method can accurately identify the concentration-time combination with the greatest reversal potential, thus providing high-quality input for subsequent cross-spectral aggregation and ultimately improving the accuracy and reliability of intervention predictions for phenotypes such as liver fibrosis.

[0076] In an exemplary embodiment, the representative value among the N reverse matching results can be determined using various mathematical calculation methods, including but not limited to: minimum value, mean, median, or weighted average.

[0077] For example, when using the minimum value, the representative value is the minimum value among N reverse matching results.

[0078] When using the mean, the representative value can be expressed as: The representative value is calculated as (Σ RGES_n) / N (Equation 3), where Σ represents summation; RGES_n represents the nth reverse match result; and N is the number of reverse match results.

[0079] When using the median, take the median of the N RGES values; When using a weighted average, different weights w_n are assigned based on the confidence level of the perturbation conditions (such as the number of sample repetitions and signal strength), and the representative value can be expressed as: The representative value is calculated as Σ (w_n × RGES_n) / Σ w_n (Equation 4), where Σ represents summation; w_n represents the weight of the nth reverse matching result; and RGES_n represents the nth reverse matching result.

[0080] This embodiment provides multiple representative value calculation methods to enhance the flexibility of the approach. Different calculation methods are suitable for different data scenarios: for example, when the quality of perturbation conditions varies, weighted averaging can assign greater weight to high-confidence conditions; when the most conservative estimate is required, the median can be used to resist the influence of extreme values. This allows for the selection of the optimal aggregation strategy based on specific application scenarios (such as high-throughput drug screening or verification of active ingredients in traditional Chinese medicine), thereby maintaining the strength of the reversal signal while further reducing noise interference and improving the robustness of the prediction results.

[0081] In a preferred embodiment of this application, the minimum value (i.e., the most negative value) among N reverse matching results is used as the representative value to obtain the target result. Using the minimum value as the representative value implements the strategy of "taking the strongest reversal signal under each phenotypic feature spectrum," which can automatically screen out the experimental conditions under which the intervention performs best among multiple perturbation conditions, thereby maximizing the capture of reversal potential. Even if some perturbation conditions (such as optimal concentration or time point) produce weak signals, the overall evaluation results will not be diluted. This effectively solves the problem that the "representative value" may be lowered by weak signals due to the average or median. In actual verification (such as testing on 50 known approved drug diseases), it significantly increases the proportion of effective interventions in the Top N candidates (82% in this application method, better than the 68% of traditional methods), thereby greatly improving screening efficiency and the recall rate of effective interventions.

[0082] In an exemplary embodiment, the method further includes a step of standardizing the inverse matching results to obtain standard inverse matching results. This step is typically performed after all inverse matching results (RGES values) for each m-th feature spectrum have been calculated and before the target result is selected, or it can be performed uniformly after all interventions and the original RGES of all feature spectra have been calculated.

[0083] Specifically, the standardization process can employ the Min-Max method, mapping all original RGES values ​​under the same predetermined phenotype to the interval [-1, 1]. The specific calculation formula is as follows: V_{s,j} = 2 × (v_{s,j} - v_{min,s}) / (v_{max,s} - v_{min,s}) - 1, equation 5); in: v_{s,j} is the original RGES value of the j-th reverse matching result under the s-th feature spectrum; v_{min,s} and v_{max,s} are the minimum and maximum original RGES values ​​of all interventions (or all calculated RGES values) under the s-th feature spectrum, respectively. V_{s,j} represents the standardized inverse matching result, with a value range of [-1, 1].

[0084] In practice, firstly, all M feature spectra of the same predetermined phenotype are statistically analyzed: RGES values ​​generated by all N sets of perturbation conditions of all K interventions under the phenotype on the spectrum are traversed to determine the global v_min,s and v_max,s of the spectrum; then, the above formula is applied to each RGES value for linear transformation.

[0085] For example, suppose that for a feature spectrum (sth feature) of breast cancer phenotype, the calculated RGES values ​​of all interventions under this spectrum are in the range of [-0.8, 0.6], i.e. v_min,s = -0.8, v_max,s = 0.6.

[0086] For a given perturbation condition of an intervention, with its original RGES = -0.5, then after standardization: V = 2 × (-0.5 - (-0.8)) / (0.6 - (-0.8)) - 1 = 2 × (0.3) / 1.4 - 1≈ 2 × 0.214 - 1 ≈ -0.571 The standardized value is closer to -1, indicating that the perturbation condition has a strong reversal potential under this spectrum.

[0087] This embodiment eliminates the numerical scale inconsistency caused by differences in sample size, batch effect, or enrichment score distribution between different feature spectra through Min-Max normalization, ensuring that the inverse matching results on M feature spectra from different sources have a unified scale that can be directly compared. In subsequent selection of target results (representative values) and cross-spectral aggregation processing, all signals are calculated under the same dimension, avoiding weight imbalance or signal submersion caused by significant differences in the original score scales.

[0088] In practical applications, such as screening candidate drugs for type 2 diabetes, different feature spectra may originate from blood samples, pancreatic tissue, or animal models, resulting in significant differences in the distribution range of the original RGES values ​​(some spectra have RGES concentrated in [-0.3, 0.2], while others are concentrated in [-0.9, 0.5]). Direct aggregation without standardization may lead to the low-scale spectrum signal being dominated by the high-scale spectrum. Through the standardization process in this step, the inverted signals of all spectra are fairly mapped to [-1, 1], thus enabling the representative result (rRGES) obtained from aggregation to truly reflect the comprehensive inverting ability of the intervention on the entire phenotypic multi-source data. This significantly improves the stability of the prediction results and the transferability across datasets, ultimately making the prioritization of candidate interventions more reliable and biologically meaningful.

[0089] It is understandable that those skilled in the art may also choose other standardization methods (such as Z-score standardization) as alternatives based on the characteristics of the data.

[0090] In an exemplary embodiment, a standard target result can also be determined based on the standard inverse matching results. Similarly, the standard target result can also be determined by performing mathematical calculations on N standard inverse matching results. These mathematical calculations include, but are not limited to, minimum, mean, median, or weighted average. Specific calculation methods are described above and will not be repeated here.

[0091] In an exemplary preferred embodiment, the minimum value among the above N standard reverse matching results is used as the representative value to obtain the standard target result.

[0092] S130, the M target results of the k-th intervention are aggregated to obtain the representative results of the k-th intervention, wherein the M target results are obtained by traversing M whole-genome gene expression feature profiles; This step involves aggregating the M target results obtained after traversing M genome-wide gene expression profiles for the k-th intervention to generate a single comprehensive score (i.e., representative result rRGES) for the current predetermined phenotype, thereby completing the transformation from multi-source heterogeneous data to final priority ranking.

[0093] In exemplary embodiments, the aggregation process can be implemented in various ways to adapt to different data characteristics and application requirements. For example, the arithmetic mean of the M target results can be taken directly, the median can be used to resist interference from individual abnormal spectra, different weights can be assigned to each feature spectrum based on the sample size or statistical significance for weighted averaging, or even only the strongest signal can be retained before aggregation. Those skilled in the art can dynamically select the aggregation strategy based on the degree of heterogeneity of the predetermined phenotype or the number of perturbation conditions from the intervention.

[0094] In an exemplary preferred embodiment, an averaging aggregation method is used to process the M target results of the k-th intervention. Specifically, firstly, negative results are selected from the M target results (positive or zero values ​​are not included in the aggregation to ensure that only signals with true reversal potential are included). Then, the arithmetic mean of these negative target results is calculated to obtain the representative result rRGES of the intervention for the current predetermined phenotype. This averaging aggregation process is essentially an equal-weighted integration of the "strongest reversal signal" under multiple feature spectra, and the formula can be expressed as: rRGES_k = (1 / N_d) × Σ min_V_s, equation 6); Where N_d is the number of feature spectra that produce negative target results, and min_V_s is the minimum inverse matching result after standardization under the s-th feature spectrum.

[0095] By employing an average aggregation method, reversal signals originally scattered across M independent feature profiles (from different studies, organizations, and platforms) are organically integrated into a stable and reliable comprehensive score. Related technologies rely on a single phenotypic gene set for one-time calculations; if the selected gene set is not representative or affected by batch effects, the prediction results can fluctuate drastically or even miss effective interventions. This embodiment, however, effectively mitigates noise, sample heterogeneity, and experimental bias from a single data source through average aggregation across M whole-genome feature profiles, ensuring that the final representative results are no longer influenced by the randomness of individual feature profiles. In practical applications, such as screening candidate drugs for breast cancer, even if a feature profile from a blood sample produces a weak signal due to patient subtype differences, strong reversal signals from other feature profiles from tumor tissues or cell line models can still be retained and contribute weight through average aggregation. This allows the rRGES value to more accurately reflect the drug's reversal potential for the overall breast cancer phenotype, significantly improving the stability of the prediction results and the recall rate of effective candidates.

[0096] Furthermore, when the M value is large (e.g., the same disease in the database of this application can have dozens of feature spectra), the average aggregation can also naturally reflect the statistical "majority voting" effect, further enhancing the robustness of the method. In scenarios such as high-throughput screening of active ingredients in traditional Chinese medicine or evaluation of pesticide disease-resistant compounds, this embodiment can handle massive amounts of heterogeneous data without causing overall ranking distortion due to bias from any single source, ultimately providing users with a more reliable priority list of candidate interventions.

[0097] S140, sort the representative results of the K interventions to determine the target intervention, wherein the representative results of the K interventions are obtained by traversing the K interventions.

[0098] This step involves globally prioritizing all K candidate interventions based on the representative results (rRGES) obtained after aggregating the multi-source phenotypic feature spectra of each intervention, thereby identifying the most promising target intervention for the current predetermined phenotype.

[0099] In an exemplary embodiment, all representative results rRGES_k (k=1,2,...,K) calculated after traversing K interventions are collected. Each rRGES_k originates from the aforementioned steps: the strongest reversal signal (target result) extracted from the expression feature profiles of M whole-genome genes for each intervention, and then aggregated (preferably average aggregation) to obtain a comprehensive score. The numerical characteristics of the representative results are: the smaller the value (the more negative), the stronger the reversal potential of the intervention for the predetermined phenotype.

[0100] The sorting process uses an ascending order, ranking the K interventions according to their rRGES values ​​from smallest to largest (from most negative to more positive). Sorting can be performed directly in memory using a stable sorting algorithm (such as Python's `sorted` function or `numpy.argsort`), and supports additional metadata (such as intervention ID, compound name, molecular structure, etc.) for subsequent output and visualization. In actual calculations, if some interventions have positive or near-zero rRGES values ​​(with no significant reversal potential), they can be placed at the end of the sort or marked as low priority as needed.

[0101] After ranking, the top-ranked interventions are identified as target interventions. Typically, a threshold is set or a fixed Top N is established based on application requirements. For example, interventions with rRGES < -0.5 are selected as high-potential candidates, or a Top 100 or Top 500 list is directly output as the final screening result. In high-throughput scenarios, the ranking results can be further refined by combining other filtering conditions (such as known toxicity, drug similarity, and structural druggability).

[0102] In practical applications, such as screening candidate drugs for breast cancer phenotypes, even if a drug shows weak reversal signals in individual feature spectra (e.g., blood-derived), if it exhibits strong reversal in multiple spectra from tumor tissue-derived sources, the aggregated rRGES value can still place it at the top, avoiding the problem of missed effective drugs due to data limitations in traditional single-query models. In the modernization research of traditional Chinese medicine, when ranking hundreds of active ingredients of traditional Chinese medicine, this step can quickly output a Top 50 candidate list, providing efficient priority guidance for subsequent in vitro / in vivo validation. In the agricultural field, when screening new pesticides or growth regulators, it can directly output a Top candidate compound list targeting specific plant disease phenotypes, improving the efficiency and reliability of identifying high-potential interventions from a massive number of compounds.

[0103] Through the above sorting and determination mechanism, this embodiment realizes a complete closed loop from raw multi-source data to the final priority list of candidate interventions, which significantly improves the interpretability and practical value of the prediction results and provides a more reliable computational screening tool for multiple fields such as drug repositioning, modernization of traditional Chinese medicine, toxicity assessment and agricultural protection.

[0104] The above text combined Figure 1 This paper describes an embodiment of the target intervention method for determining the present application. The following section, in conjunction with… Figure 2 An embodiment of the apparatus described in this application is presented.

[0105] Figure 2 This is a schematic block diagram of a target intervention determination device 200 provided in an embodiment of this application, wherein the device 200 can be configured in an electronic device.

[0106] refer to Figure 2 The target intervention determination device 200 includes: a signal acquisition module 210, a reverse matching module 220, an aggregation module 230, and a decision module 240.

[0107] The system includes a signal acquisition module 210, which acquires N sets of gene expression feature sets for K interventions and M whole-genome gene expression feature profiles corresponding to a predetermined phenotype from a reference database, where K, N, and M are all integers not less than 1; a reverse matching module 220, which, based on mathematical transformation, determines the reverse matching results of the N sets of gene expression feature sets relative to the m-th whole-genome gene expression feature profile for the k-th intervention to obtain a target result, where the target result is a representative value of the reverse matching result, k is an integer not less than 1 and not greater than K, and m is an integer not less than 1 and not greater than M; an aggregation module 230, which aggregates the M target results of the k-th intervention to obtain a representative result of the k-th intervention, where the M target results are obtained by traversing the M whole-genome gene expression feature profiles; and a decision module 240, which sorts the representative results of the K interventions to determine the target intervention, where the representative results of the K interventions are obtained by traversing the K interventions.

[0108] In an exemplary embodiment, the target result is a negative value.

[0109] In an exemplary embodiment, the step of determining the reverse matching results of N gene expression feature sets relative to the m-th whole-genome gene expression feature spectrum based on mathematical transformation for the k-th intervention to obtain the target result includes: dividing the n-th gene expression feature set to obtain the n-th upregulated gene set and the n-th downregulated gene set, where n is an integer not less than 1 and not greater than N; determining the enrichment scores of the n-th upregulated gene set and the n-th downregulated gene set respectively; determining the reverse matching results of the n-th gene expression feature set relative to the m-th whole-genome gene expression feature spectrum based on the difference between the enrichment scores of the n-th upregulated gene set and the n-th downregulated gene set; traversing the N gene expression feature sets to obtain N reverse matching results; and determining the representative value among the N reverse matching results to obtain the target result.

[0110] In an exemplary embodiment, determining the representative value among the N reverse matching results to obtain the target result includes: performing mathematical calculations on the N reverse matching results to determine the representative value among the N reverse matching results to obtain the target result. The mathematical calculations include: minimum value, mean, median, or weighted average.

[0111] In an exemplary embodiment, determining the representative value among the N reverse matching results to obtain the target result includes: using the minimum value among the N reverse matching results as the representative value to obtain the target result.

[0112] In an exemplary embodiment, the method further includes: standardizing the reverse matching result to obtain a standard reverse matching result, wherein the standardization process includes Min-Max standardization.

[0113] In an exemplary embodiment, the aggregation process of the M target results of the k-th intervention to obtain the representative result of the k-th intervention includes: averaging the M target results of the k-th intervention to obtain the representative result of the k-th intervention.

[0114] In an exemplary embodiment, the reference database is constructed as follows: the raw transcriptome sequencing data is subjected to quality control processing to obtain high-quality sequencing reads; the high-quality sequencing reads are aligned to a reference genome to generate a comparison file; based on the comparison file, the expression level of each gene in each sample is estimated; differential expression analysis is performed on the expression level of each gene in each sample in the experimental group and the control group to determine the whole-genome gene expression profile, wherein the whole-genome gene expression profile includes the differential expression fold change and the correction p-value for each gene.

[0115] In an exemplary embodiment, the N sets of gene expression feature sets are upregulated gene sets and downregulated gene sets induced by the intervention under different perturbation conditions, wherein the perturbation conditions include at least one of: cell line, treatment concentration, and treatment time.

[0116] In an exemplary embodiment, the intervention includes: a compound, the compound including at least one of a drug, a pesticide, and a growth regulator; the predetermined phenotype includes: a disease.

[0117] It should be understood that the device embodiments and method embodiments can correspond to each other, and similar descriptions can be referred to the method embodiments. To avoid repetition, further details will not be provided here. Specifically, Figure 2 The device 200 shown can perform Figure 1 The corresponding method embodiments, and the foregoing and other operations and / or functions of each module in the device 200 are respectively implemented to achieve Figure 1 For the sake of brevity, the corresponding processes in each method are not described in detail here.

[0118] The apparatus 200 of this application embodiment has been described above from the perspective of functional modules in conjunction with the accompanying drawings. It should be understood that this functional module can be implemented in hardware, in software instructions, or in a combination of hardware and software modules. Specifically, the steps of the method embodiments in this application can be completed by integrated logic circuits in the processor's hardware and / or by software instructions. The steps of the method disclosed in this application embodiment can be directly embodied as being executed by a hardware decoding processor, or by a combination of hardware and software modules in the decoding processor. Optionally, the software module can be located in a mature storage medium in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, etc. This storage medium is located in memory, and the processor reads information from the memory and, in conjunction with its hardware, completes the steps in the above method embodiments.

[0119] Figure 3 This is a schematic block diagram of the electronic device 300 provided in an embodiment of this application. The electronic device 300 may be the aforementioned training device or execution device, but is not limited thereto. Figure 3 As shown, the electronic device 300 may include: The system includes a memory 310 and a processor 320. The memory 310 stores a computer program 330 and transfers the computer program 330 to the processor 320. In other words, the processor 320 can retrieve and run the computer program 330 from the memory 310 to implement the methods described in the embodiments of this application.

[0120] For example, the processor 320 can be used to execute the steps in the above method according to the instructions in the computer program 330.

[0121] In some embodiments of this application, the processor 320 may include, but is not limited to: General-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc.

[0122] In some embodiments of this application, the memory 310 includes, but is not limited to: Volatile memory and / or non-volatile memory. Non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. Volatile memory can be random access memory (RAM), used as an external cache. By way of example, but not limitation, many forms of RAM are available, such as Static RAM (SRAM), Dynamic RAM (DRAM), Synchronous DRAM (SDRAM), Double Data Rate SDRAM (DDR SDRAM), Enhanced SDRAM (ESDRAM), Synchronous Link DRAM (SLDRAM), and Direct Rambus RAM (DR RAM).

[0123] In some embodiments of this application, the computer program 330 may be divided into one or more modules, which are stored in the memory 310 and executed by the processor 320 to perform the method provided in this application. The one or more modules may be a series of computer program instruction segments capable of performing specific functions, which describe the execution process of the computer program 330 in the electronic device.

[0124] like Figure 3 As shown, the electronic device 300 may further include: Transceiver 340, which can be connected to processor 320 or memory 310.

[0125] The processor 320 can control the transceiver 340 to communicate with other devices; specifically, it can send information or data to other devices or receive information or data sent by other devices. The transceiver 340 may include a transmitter and a receiver. The transceiver 340 may further include antennas, and the number of antennas may be one or more.

[0126] It should be understood that the various components in the electronic device 300 are connected through a bus system, which includes a data bus, a power bus, a control bus, and a status signal bus.

[0127] According to one aspect of this application, a computer-readable storage medium is provided that stores computer instructions or programs thereon, which, when executed by a computer, enable the computer to perform the methods of the above-described method embodiments. Alternatively, embodiments of this application also provide a computer program product containing instructions that, when executed by a computer, cause the computer to perform the methods of the above-described method embodiments.

[0128] According to another aspect of this application, a computer program product or computer program is provided, comprising computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the method described in the above-described method embodiments.

[0129] In other words, when implemented using software, it can be implemented wholly or partially in the form of a computer program product. This computer program product includes one or more computer instructions. When these computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., digital video disc (DVD)), or a semiconductor medium (e.g., solid-state disk (SSD)).

[0130] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0131] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or modules may be electrical, mechanical, or other forms.

[0132] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to implement the solution of this embodiment according to actual needs. For example, the functional modules in the various embodiments of this application may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module.

[0133] Embodiments of this application will now be described in more detail, examples of which are illustrated in the accompanying drawings. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application.

[0134] Example 1: Validation of the method for determining the target intervention To verify the effectiveness, accuracy, and robustness of the proposed solution, it is compared with the L1000-based CMap method, as follows: Fifty diseases with known approved therapeutic drugs were selected as test phenotypes. Multiple genome-wide gene expression profiles (M≥3, from different organizations and studies) corresponding to each disease were extracted from a reference database. Simultaneously, gene expression profiles of known small molecule compounds were obtained from the L1000FWD database as input for candidate interventions.

[0135] The representative result rRGES for each compound was calculated using the method described in this application (steps S110-S140), and then sorted in ascending order. It was determined whether any known approved treatments for the disease were included in the Top 200. The results are as follows: Figure 4 As shown: the horizontal axis represents the compound ranking position (Top 1 to Top 200), and the vertical axis represents the proportion of diseases containing at least one known therapeutic agent within the corresponding Top range. The method of this application (orange curve) identifies an effective drug in 80% of the diseases within the Top 200, significantly higher than the 66% of the traditional CMap method (blue curve). The green curve represents the performance of the combined results of the two methods, showing a clear complementary effect.

[0136] Further analysis of the complementarity of the two methods: such as Figure 5 As shown, the vertical axis represents the number of known therapeutic drugs found. Within the Top 40, the number of drugs found by combining the results of the two methods is close to the sum of the numbers found by each method alone, indicating extremely low overlap. Even when extended to the Top 200, the number of overlapping drugs remains small. This demonstrates that the "intervention-based query for predetermined phenotype" reverse query paradigm proposed in this application is highly complementary to the traditional "intervention-based query for predetermined phenotype" model in terms of prediction results, significantly expanding the candidate drug pool and improving the coverage of drug relocation.

[0137] The experimental results above demonstrate that this application, through the whole transcriptome multi-source feature spectrum database and the representative result rRGES framework (optimal reversal signal screening for each spectrum + cross-spectrum averaging aggregation), effectively overcomes the problems of information loss, unstable results, and missed screening caused by incomplete gene coverage and single phenotypic characterization in related technologies, and achieves significant improvements in accuracy, robustness, and candidate coverage.

[0138] Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of this application.

Claims

1. A method for determining a target intervention, characterized in that, include: Obtain N sets of gene expression feature sets for K interventions, and obtain M whole-genome gene expression feature profiles corresponding to a predetermined phenotype from a reference database, where K, N, and M are all integers not less than 1; For the k-th intervention, based on mathematical transformation, the reverse matching results of N groups of gene expression feature sets relative to the m-th whole genome gene expression feature spectrum are determined to obtain the target result, wherein the target result is the representative value of the reverse matching result, k is an integer not less than 1 and not greater than K, and m is an integer not less than 1 and not greater than M. The M target results of the k-th intervention are aggregated to obtain the representative results of the k-th intervention, wherein the M target results are obtained by traversing M whole-genome gene expression feature profiles; The representative results of the K interventions are sorted to determine the target intervention, wherein the representative results of the K interventions are obtained by traversing the K interventions.

2. The method according to claim 1, characterized in that, The target result is a negative value.

3. The method according to claim 2, characterized in that, For the k-th intervention, based on mathematical transformations, the reverse matching results of N groups of gene expression feature sets relative to the m-th whole-genome gene expression feature spectrum are determined to obtain the target results, including: The expression feature set of the nth gene is divided into the nth upregulated gene set and the nth downregulated gene set, where n is an integer not less than 1 and not greater than N; Determine the enrichment scores of the nth upregulated gene set and the nth downregulated gene set, respectively; Based on the difference between the enrichment scores of the nth upregulated gene set and the nth downregulated gene set, the reverse matching result of the nth gene expression feature set relative to the mth whole genome gene expression feature profile is determined; Traverse N sets of gene expression feature sets to obtain N reverse matching results; The representative value among the N reverse matching results is determined to obtain the target result.

4. The method according to claim 3, characterized in that, Determining the representative value among the N reverse matching results to obtain the target result includes: Mathematical calculations are performed on the N reverse matching results to determine the representative value among the N reverse matching results, and the target result is obtained. The mathematical calculations include: minimum value, mean, median or weighted average.

5. The method according to claim 4, characterized in that, Determining the representative value among the N reverse matching results to obtain the target result includes: The minimum value among the N reverse matching results is used as the representative value to obtain the target result.

6. The method according to claim 1, characterized in that, Further includes: The reverse matching results are standardized to obtain standard reverse matching results, wherein the standardization process includes Min-Max standardization.

7. The method according to any one of claims 1-6, characterized in that, The aggregation process of the M target results of the k-th intervention to obtain the representative result of the k-th intervention includes: The representative result of the k-th intervention is obtained by averaging and aggregating the M target results of the k-th intervention.

8. The method according to any one of claims 1-6, characterized in that, The reference database is constructed as follows: The raw transcriptome sequencing data were subjected to quality control processing to obtain high-quality sequencing reads; The high-quality sequencing reads are aligned to a reference genome to generate a comparison file; Based on the comparison file, the expression level of each gene in each sample is estimated; Differential expression analysis was performed on the expression levels of each gene in each sample from the experimental and control groups to determine the genome-wide gene expression profile, which includes the differential expression fold and corrected P-value for each gene.

9. The method according to any one of claims 1-6, characterized in that, The N sets of gene expression feature sets are the upregulated gene sets and downregulated gene sets induced by the intervention under different perturbation conditions, wherein the perturbation conditions include at least one of the following: cell line, treatment concentration, and treatment time.

10. The method according to any one of claims 1-6, characterized in that, The intervention includes: a compound, the compound including at least one of a drug, a pesticide, and a growth regulator; the predetermined phenotype includes: a disease.

11. A target intervention object determination device, characterized in that, include: The signal acquisition module is used to acquire N sets of gene expression feature sets of K interventions and to acquire M whole-genome gene expression feature profiles corresponding to a predetermined phenotype from a reference database, where K, N, and M are all integers not less than 1; The reverse matching module is used to determine the reverse matching results of N sets of gene expression feature sets relative to the m-th whole genome gene expression feature spectrum for the k-th intervention based on mathematical transformation, so as to obtain the target result. The target result is the representative value of the reverse matching result, k is an integer not less than 1 and not greater than K, and m is an integer not less than 1 and not greater than M. An aggregation module is used to aggregate the M target results of the k-th intervention to obtain the representative results of the k-th intervention, wherein the M target results are obtained by traversing M whole-genome gene expression feature profiles; The decision module is used to sort the representative results of K interventions and determine the target intervention, wherein the representative results of the K interventions are obtained by traversing the K interventions.

12. An electronic device, characterized in that, include: Processor and memory; The memory is used to store computer programs; The processor is configured to execute the computer program to implement the target intervention determination method as described in any one of claims 1 to 10.

13. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions or programs that, when executed on a computer, cause the target intervention determination method as described in any one of claims 1 to 10 to be performed.

14. A computer program product, characterized in that, The computer program product includes computer instructions that, when some or all of the computer instructions are run on a computer, cause the target intervention determination method as described in any one of claims 1 to 10 to be executed.

Images