An algorithm for traditional Chinese medicine compound modification and prediction

Abstract

This invention belongs to the interdisciplinary field of bioinformatics and traditional Chinese medicine (TCM), and discloses an algorithm for the improvement and prediction of TCM compound prescriptions, solving the problems of strong subjectivity, long cycle, and low efficiency in traditional TCM compound prescription improvement. The algorithm first acquires and preprocesses omics sequencing data from disease and healthy groups, then establishes TCM-target gene associations based on the HIT2 and TCMSP databases, and constructs a three-dimensional association matrix of TCM-target genes-5hmC loci for epigenomic data. After dimensionality reduction and extraction of core principal components using PCA, the sample-TCM association score is calculated and a matrix is ​​constructed. The p-value is obtained through t-test or Wilcoxon rank-sum test, and converted into DAS values ​​to quantify the regulatory activity of TCM, thus screening TCM for compound prescription improvement. This invention's algorithm is objective, highly targeted, efficient, and widely applicable, adaptable to various omics data and disease scenarios. Its effectiveness has been verified in a coronary atherosclerosis case, providing precise technical support for TCM compound prescription screening and optimization, personalized TCM treatment, and the development of innovative TCM drugs.

Description

Technical Field

[0001] This invention relates to the interdisciplinary field of bioinformatics and traditional Chinese medicine (TCM) technology, specifically to an algorithm for improving and predicting TCM compound prescriptions. It is applicable to the evaluation of TCM activity differences and the screening and optimization of TCM compound prescriptions based on various sequencing omics data. It can also be extended to scenarios such as the modern interpretation of TCM, the formulation of personalized TCM treatment plans, and the research and development of innovative TCM drugs. Background Technology

[0002] The efficacy of traditional Chinese medicine (TCM) compound prescriptions relies on the synergistic effects of multiple components and targets. However, improvements to traditional TCM compound prescriptions often depend on clinical experience, which can lead to issues such as high subjectivity, long processing times, and low efficiency. With the rapid development of omics technologies, various omics sequencing data (including but not limited to genomics, transcriptomics, epigenomics, proteomics, and metabolomics) are increasingly widely used in disease diagnosis and drug development. The biomarkers they contain can comprehensively reflect the molecular mechanisms of disease development and progression, providing rich quantitative evidence for objectively evaluating drug efficacy.

[0003] Currently, there is no mature algorithm that can systematically link various omics sequencing data with traditional Chinese medicine (TCM) targets and achieve TCM compound improvement and efficacy prediction through standardized quantitative analysis. Most TCM activity assessment methods rely only on single targets or phenotypic indicators, making it difficult to comprehensively capture the overall regulatory effects of multiple TCM components on the disease molecular network, and also failing to provide accurate and universally applicable quantitative evidence for compound improvement. Therefore, there is an urgent need to develop an objective and efficient algorithm for TCM compound improvement and prediction that is compatible with various omics data, overcoming the technical bottlenecks of traditional methods, such as strong subjectivity, insufficient targeting, and limited applicability. Summary of the Invention

[0004] This invention aims to provide an algorithm for the improvement and prediction of traditional Chinese medicine (TCM) compound prescriptions. By integrating sequencing omics data with TCM target information, it quantitatively evaluates the activity differences of different TCMs between disease groups and healthy groups, providing objective and accurate technical support for the screening, optimization and improvement of TCM compound prescriptions.

[0005] The core technical solution of this invention is achieved through the following steps, specifically including:

[0006] I) Omics Data Acquisition and Preprocessing: Omics data of samples, including disease group samples and healthy group samples, were acquired through sequencing, downloading from public databases, etc. Sequencing data underwent quality control, low-quality site filtering, and standardization to construct the sample expression matrix. Subsequently, variance stabilization transformation (VST) was performed on the overall data using the DEseq2 package in R language.

[0007]

[0008] II) Association between Traditional Chinese Medicine (TCM) and target genes: Based on various TCM target databases (such as HIT2, TCMSP), target genes of various TCMs are obtained. For epigenomic data, further extension from target genes to epigenetic modification sites is needed. Ultimately, a precise association between TCM, genes, and epigenetic modification sites will be achieved.

[0009] III) PCA dimensionality reduction and core component extraction: targeting the set of target features for a single Chinese herbal medicine h. Sample set Construct a normalized matrix containing only the target. As the input matrix, the dimensionality is reduced by PCA while preserving the cumulative variance contribution rate. 70% of the front One principal component is obtained, and the PC1 load vector is extracted. The contribution of the 5hmC site to the direction of maximum variation is represented.

[0010] IV) Scoring Calculation at the Sample Level for Traditional Chinese Medicine: For the sample For each Chinese medicine (TCM) herb (h), the weighted expression value of its target set is calculated, and the sample-TCM score matrix is ​​obtained. (i samples, H kinds of Chinese medicine):

[0011]

[0012] V) Calculation of Differences in Traditional Chinese Medicine Activity: DAS compares the sample-TCM score distribution (sample group label) between the healthy control group (HC) and the disease group (DC). ), quantifying the regulatory effect of traditional Chinese medicine (TCM) between the two groups. First, the score vectors of TCM h were extracted from the HC and DC groups: , An independent two-sample t-test was performed using equation (3), and the two-sided p-value was calculated from the t-distribution. Then, it is converted to DAS using equation (4). The calculation formula is as follows:

[0013]

[0014]

[0015] The algorithm provided by this invention is based on omics data preprocessing, uses the association between traditional Chinese medicine and target genes as a bridge, PCA dimensionality reduction to extract core features as the core, and quantitative scoring to verify differences as a means to form a complete traditional Chinese medicine compound improvement technology system. It is adaptable to multiple types of omics data and disease scenarios, and achieves precision and efficiency in compound improvement by replacing experience judgment with objective data.

[0016] Compared with the prior art, the present invention has the following beneficial effects:

[0017] 1. Strong objectivity: Based on the precise correlation between sequencing omics data and TCM targets, it avoids the experience-based dependence of traditional TCM compound improvement and provides an objective quantitative basis for TCM activity evaluation.

[0018] 2. Highly targeted: By using PCA to extract core differential components and focusing on the specific targets of traditional Chinese medicine, the regulatory effects of traditional Chinese medicine on disease-related genes can be accurately captured.

[0019] 3. High efficiency: It can simultaneously evaluate the activity differences of multiple Chinese medicines in batches, quickly screen out Chinese medicine components with significant disease regulation potential, and greatly shorten the improvement cycle of Chinese medicine compound prescriptions.

[0020] 4. Wide applicability: It can be extended to the improvement of traditional Chinese medicine compound prescriptions for various disease types. It only requires replacing the corresponding omics sequencing samples for the disease, and is not limited to a certain type of omics data. It has good versatility and scalability. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments of this application will be briefly introduced below.

[0022] Figure 1 This is a schematic diagram illustrating the treatment of macrophages with a single herb and the detection of CD36 expression levels in vitro.

[0023] Figure 2 A schematic diagram for detecting the degree of atherosclerosis in the aorta of mice;

[0024] Figure 3 This diagram illustrates the Oil Red-O staining area and the proportion of foam cells in different groups of mice. Detailed Implementation

[0025] The embodiments of this application are described in detail below. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain this application, and should not be construed as limiting this application.

[0026] Those skilled in the art will understand that, unless specifically stated otherwise, the singular forms “a,” “an,” and “the” used herein may also include the plural forms. It should be further understood that the term “comprising” as used in the specification of this application means the presence of features, integers, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof. The term “and / or” as used herein includes all or any units and all combinations of one or more associated listed items.

[0027] To make the objectives, technical solutions, and advantages of this application clearer, the specific embodiments of this application will be described in further detail below with reference to the accompanying drawings.

[0028] The core of this invention is to provide an algorithm for improving and predicting traditional Chinese medicine compound prescriptions, comprising the following steps:

[0029] Step 1: Obtain omics sequencing data from the disease group and the healthy group, and preprocess the omics sequencing data to construct a sample-site expression matrix.

[0030] Preferably, the preprocessing of the omics sequencing data includes: filtering low-quality sequencing sites, cleaning and standardizing the retained sites to construct a standardized expression matrix, and calling the DEseq2 package in R language to perform variance stabilization transformation on the standardized expression matrix to eliminate interference caused by differences in sequencing depth.

[0031] Preferably, the filtering of low-quality sequencing sites includes sites with sequencing quality values ​​Q < 30.

[0032] Specifically, raw sequencing data from the disease group and the healthy group (such as plasma 5hmC-Seal sequencing data from coronary atherosclerosis cases) are obtained. Acquisition methods include sequencing and downloading from public databases to obtain omics data from the samples. The samples include disease group samples and healthy group samples. Basic quality control preprocessing operations are performed, including filtering low-quality sites, specifying sites with Q < 30 as the filtering threshold, and retaining high-quality sequencing sites; subsequently, data cleaning is performed on the retained sites to remove duplicates and outliers, completing data standardization. Constructing the sample-site expression matrix includes: building the standardized expression matrix using R language; finally, calling the DEseq2 package in R language to perform variance stabilization transformation (VST) on the overall data.

[0033]

[0034] Variance stabilization transformation of the expression matrix eliminates interference from sequencing depth differences, providing a high-quality, unbiased omics data foundation for subsequent analysis.

[0035] Step 2: Establish the association between traditional Chinese medicine and target genes based on the traditional Chinese medicine target database, and establish a multidimensional mapping relationship between traditional Chinese medicine, target genes and 5hmC sites based on epigenomic data to obtain the set of specific sites corresponding to each traditional Chinese medicine.

[0036] Preferably, the traditional Chinese medicine target database is the HIT2 database and / or the TCMSP database. The construction of the multidimensional mapping relationship includes: retrieving the target genes of each traditional Chinese medicine to be evaluated, forming a basic association table of traditional Chinese medicine-target genes, matching the 5hmC sites of the corresponding target genes, and integrating them to obtain a standardized three-dimensional association matrix of traditional Chinese medicine-target genes-5hmC sites.

[0037] Specifically, based on authoritative TCM target databases (HIT2, TCMSP, etc.), known target genes for each TCM to be evaluated were retrieved, establishing associations between TCM and target genes, forming a basic association table of "TCM name - target gene number," completing the initial matching of single TCMs with core targets, and building the core framework for subsequent multidimensional associations. Based on epigenomic data, a multidimensional mapping relationship between TCM, target genes, and 5hmC sites was established to accurately pinpoint the specific molecular sites regulated by TCM, improving the accuracy of associations and obtaining the set of specific sites corresponding to each TCM.

[0038] By integrating the basic association table with the multidimensional epigenetic association table, unifying the data format (such as standardizing gene and locus numbers), and eliminating redundant association information, a three-dimensional association matrix of "traditional Chinese medicine-target gene-epiotic locus" that can be directly used for subsequent analysis is finally formed, ensuring data consistency and retrieval.

[0039] Step 3: Perform PCA dimensionality reduction on the target feature set of traditional Chinese medicine and extract the core principal components.

[0040] Preferably, the step of performing PCA dimensionality reduction on the target feature set of traditional Chinese medicine and extracting core principal components includes: constructing a two-dimensional feature matrix of traditional Chinese medicine-features or a three-dimensional feature matrix of samples-traditional Chinese medicine-features based on the association data of traditional Chinese medicine-target genes or the three-dimensional association matrix of traditional Chinese medicine-target genes-5hmC sites; performing PCA dimensionality reduction calculation on the feature matrix, and screening core principal components whose cumulative variance contribution rate is greater than or equal to a preset value.

[0041] Specifically, taking a single Chinese herbal medicine (TCM) as the unit, based on the established three-dimensional correlation matrix of "TCM-target gene-epiotic site," the feature information of all target points and epigenetic sites corresponding to each TCM is extracted to construct a two-dimensional feature matrix of "TCM-feature (target / site)." For multi-sample data (such as healthy groups and disease groups), this is further expanded to a three-dimensional feature matrix of "sample-TCM-feature," fully covering the correlation relationships of sample, TCM, and molecular features. The constructed feature matrix is ​​imported into data analysis tools (such as Python and R), and the PCA dimensionality reduction algorithm is called for calculation. The algorithm decomposes the dimensions of the feature matrix, calculates the variance contribution rate of each feature, and clarifies the explanatory power of different features for data variation. A core retention rule of ≥70% cumulative variance contribution rate was set to select principal component sets that meet this condition. Among the principal components, the PCA1 loading vector was selected first, as it carries the largest amount of information variation in the matrix, accurately focusing on the specific targets and epigenetic sites of TCM regulation and eliminating interference from irrelevant features. Finally, the core regulatory features of each TCM were extracted, which helped to identify the core molecular targets for subsequent activity evaluation.

[0042] Step 4: Calculate the association score between the sample and the traditional Chinese medicine based on the core principal components, and construct the sample-traditional Chinese medicine score matrix;

[0043] Preferably, the core principal component is a sample-TCM-specific 5hmC site matrix constructed for each TCM based on its specific 5hmC site set, and the first principal component PCA1 is extracted after PCA dimensionality reduction.

[0044] Preferably, the first principal component PCA1 loading value carries the largest amount of information variation in the feature matrix, which is used to characterize the specific target and / or 5hmC site of traditional Chinese medicine regulation.

[0045] Preferably, the association score between the sample and the traditional Chinese medicine is calculated based on the core principal components, and the sample-traditional Chinese medicine score matrix is ​​constructed by: for each sample, the score of the traditional Chinese medicine in the corresponding sample is obtained by weighted summation of the expression level of the traditional Chinese medicine-specific 5hmC site and the PCA1 loading value.

[0046] Specifically, based on the extracted PCA1 core feature vector, the feature matrix is ​​weighted and calculated to calculate the association score between each sample and each Chinese herbal medicine. The score calculation is based on the weight of the core feature and combined with the expression value of the target / site. Finally, a two-dimensional score matrix of "sample-Chinese herbal medicine" is constructed to quantify the matching strength between different samples and each Chinese herbal medicine.

[0047] Among them, for samples For each Chinese medicine (TCM) herb (h), the weighted expression value of its target set is calculated, and the sample-TCM score matrix is ​​obtained. (i samples, H kinds of Chinese medicine):

[0048]

[0049] Step 5: Perform a significance test on the scores of the healthy group and the disease group, quantify the regulatory activity of traditional Chinese medicine by scoring the difference in activity, and screen out traditional Chinese medicines that can be used for compound improvement.

[0050] Preferably, quantifying the regulatory activity of traditional Chinese medicine through activity difference scoring includes: conducting a significance test on the difference in traditional Chinese medicine scores between the disease group and the healthy group, calculating the difference P value between the corresponding traditional Chinese medicine scores of the healthy group and the disease group using an independent two-sample t-test, and converting the obtained P value into an activity difference score.

[0051] Preferably, the significance test for the difference between the scores of the healthy group and the disease group includes at least one of the following: t-test and Wilcoxon rank-sum test.

[0052] Specifically, according to the score matrix By comparing the sample-to-traditional Chinese medicine score distributions of the healthy control group (HC) and the disease group (DC) (sample group labels) ), extract the score vectors of the corresponding Chinese medicine h from the healthy control group (HC) and the disease group (DC), respectively. , The significance of the differences was analyzed by using an independent two-sample t-test according to equation (3), and the p-value of the score difference between the healthy group and the disease group for each Chinese herbal medicine was calculated by the t-distribution. Substitute the P value obtained from the t-test into the DAS (Difference in Activity Score) formula (4) to complete the DAS value calculation for each Chinese herb. The magnitude of the DAS value directly reflects the strength of the regulatory activity of the Chinese herb on the disease. The higher the value, the more significant the difference in activity between the disease group and the healthy group, and the greater the potential as a core component for compound improvement.

[0053]

[0054]

[0055] Based on the calculated DAS values, all Chinese herbal medicines to be evaluated are sorted in descending order, and those with the highest DAS values ​​are selected to form a candidate list of Chinese herbal medicines with potential for improvement, providing direct quantitative basis for subsequent compound optimization.

[0056] If the data distribution does not conform to a normal distribution, the Wilcoxon rank-sum test can be used instead to ensure the reliability of the test results.

[0057] The algorithm provided has multi-scenario adaptability. Specifically, it solidifies the core steps of omics data preprocessing, traditional Chinese medicine-target association, PCA dimensionality reduction, and DAS scoring into a standardized process, and clarifies the input data format, tool calling specifications, and parameter setting standards for each step. For different disease types or different omics data types, only the core omics sequencing data and the corresponding traditional Chinese medicine target association relationship for the disease need to be replaced, and the adaptation can be completed without reconstructing the core technical logic.

[0058] The algorithm provided by this invention is based on omics data preprocessing, uses the association between traditional Chinese medicine and target genes as a bridge, PCA dimensionality reduction to extract core features as the core, and quantitative scoring to verify differences as a means to form a complete traditional Chinese medicine compound improvement technology system. It is adaptable to multiple types of omics data and disease scenarios, and achieves precision and efficiency in compound improvement by replacing experience judgment with objective data.

[0059] Compared with the prior art, the present invention has the following beneficial effects:

[0060] 1. Strong objectivity: Based on the precise correlation between sequencing omics data and TCM targets, it avoids the experience-based dependence of traditional TCM compound improvement and provides an objective quantitative basis for TCM activity evaluation.

[0061] 2. Highly targeted: By using PCA to extract core differential components and focusing on the specific targets of traditional Chinese medicine, the regulatory effects of traditional Chinese medicine on disease-related genes can be accurately captured.

[0062] 3. High efficiency: It can simultaneously evaluate the activity differences of multiple Chinese medicines in batches, quickly screen out Chinese medicine components with significant disease regulation potential, and greatly shorten the improvement cycle of Chinese medicine compound prescriptions.

[0063] 4. Wide applicability: It can be extended to the improvement of traditional Chinese medicine compound prescriptions for various disease types. It only requires replacing the corresponding omics sequencing samples for the disease, and is not limited to a certain type of omics data. It has good versatility and scalability.

[0064] The invention will be further described in detail below with reference to specific implementation examples:

[0065] Example 1: Coronary Atherosclerosis

[0066] Sample inclusion and exclusion criteria: Patients were assessed for vascular malformations, plaque distribution, and extent of obstruction through coronary angiography. Patients with vascular stenosis >50% were diagnosed with coronary heart disease and included in the disease cohort, while those with no vascular stenosis or stenosis <20% were included in the healthy cohort.

[0067] Sample source and quantity: Collected from Beijing Fuwai Hospital. The healthy group consisted of 180 cases; the disease group consisted of 190 cases.

[0068] Clinical data table:

[0069]

[0070] Omics data acquisition and preprocessing: Plasma samples were collected from volunteers in the disease group (coronary heart disease) and the healthy group. 5hmC sequencing data were obtained using 5hmC-Seal sequencing technology. Quality control of the sequencing data was performed (filtering sites with sequencing quality values ​​Q < 30), 5hmC sites were identified using MACS2, and 5hmC site data were quantified using bedtools. Finally, a sample-5hmC site expression matrix was constructed, and the processed expression matrix was obtained using the vst function in the DESeq2 package. Simultaneously, differential analysis was performed, and 13472 differentially expressed hydroxymethylation sites (DhMRs) were obtained by screening with a log2 fold change (log2FoldChange) > 0.25 and a p-value < 0.01 as the threshold.

[0071] Target association between traditional Chinese medicine and 5hmC site: Information on traditional Chinese medicine and its target sites was downloaded from the HIT2 traditional Chinese medicine target database and the association between 5hmC site and gene was established to establish a mapping relationship of "traditional Chinese medicine-target gene-5hmC site".

[0072] PCA dimensionality reduction and core component extraction: (1) Screening drugs from the perspective of adding drugs to the original prescription: For each Chinese medicine, extract the 5hmC sites related to the Chinese medicine, and combine them with the 5hmC sites mapped to the three Chinese medicines Sanqi, Xuejie and Hongqu in the original prescription of Chiqilin prescription to form the expression matrix 1 of sample × 5hmC sites; (2) Screening drugs from the perspective of potential sensitivity of patients: For each Chinese medicine, extract the 5hmC sites related to the Chinese medicine, and at the same time take the intersection with the previously screened DhMRs to form the expression matrix 2 of sample × differential 5hmC sites; PCA dimensionality reduction is performed using the prcomp function of R language, and the core component PCA1 of the two matrices is extracted respectively.

[0073] Score calculation at the sample level for traditional Chinese medicine: For each sample, the expression level at each site is extracted from two matrices, multiplied by the PCA1 loading value, and summed to obtain the score of the improved formula with the addition of a certain traditional Chinese medicine in each sample.

[0074] The scoring of TCM activity differences was calculated using an independent samples t-test to compare the differences in TCM scores between the disease group and the healthy group, thereby ranking all TCMs by score. A higher score for Pathway 1 indicates a more significant difference in the effect of the modified Chiqilin formula with added medication between the two groups of volunteers; a higher score for Pathway 2 indicates a more significant difference in the effect of the drug between the two groups of volunteers, suggesting potentially better therapeutic effects. The first five herbs in Pathway 1 are: Cnidium monnieri, Bergenia serratifolia, Hedyotis diffusa, Toad, and Saffron; the first five herbs in Pathway 2 are: Stephania tetrandra, Alpinia oxyphylla, Scoparia scabra, Solanum nigrum root, and Terminalia chebula. After screening by TCM physicians, Bergenia serratifolia and Hedyotis diffusa are not closely related to cardiovascular and cerebrovascular diseases in TCM pharmacology and there is no relevant research literature; Toad is not a drug on the Chinese Pharmacopoeia white list. Therefore, Cnidium monnieri and Saffron were selected from Pathway 1; Stephania tetrandra and Alpinia oxyphylla were selected from Pathway 2.

[0075] This embodiment successfully screened out four Chinese herbal ingredients with significant differences in activity—saffron, Cnidium monnieri, Stephania tetrandra, and Alpinia oxyphylla—using the algorithm of this invention, providing a precise basis for the improvement of the Chiqilin formula.

[0076] Example 2: In vitro macrophage detection

[0077] See Figure 1 Data showed that single-drug treatment of macrophages in vitro and detection of CD36 expression levels revealed a significant correlation between drug ranking and CD36 expression levels, indicating that the drug screening algorithm provided in this application has strong reliability.

[0078] The in vitro experimental verification steps include: selecting candidate traditional Chinese medicines screened by the algorithm, processing in vitro cell models related to the target disease respectively; detecting the expression levels of core regulatory genes in the cell models, comparing the gene expression level ranking obtained from the experiment with the DAS value ranking of traditional Chinese medicines given by the algorithm, and verifying the significant correlation between the two to confirm the reliability of the algorithm.

[0079] Table 1. Relative expression levels of CD36 in macrophages of each group in vitro (mean ± SD, n=6)

[0080]

[0081] As shown in Table 1, the ranking of the activity difference scores (DAS values) of various Chinese herbal medicines by the algorithm of this invention is significantly correlated with the ranking of the measured CD36 expression levels in in vitro macrophages. Cnidium monnieri (27.43±2.67) and Chrysanthemum morifolium (29.75±2.75), which have higher algorithm scores, can significantly reduce the CD36 expression level in macrophages to a level close to that of the control group (1.32±0.87). Conversely, Tetrapanax papyriferus (95.95±13.15) and Perilla frutescens (82.95±8.08), which have lower algorithm scores, do not show significant inhibitory effects on CD36 expression and remain close to the model group (98.23±7.14).

[0082] The above results show that the algorithm of the present invention, based on the systematic association between omics sequencing data and traditional Chinese medicine targets, can objectively and accurately reflect the actual regulatory effect of traditional Chinese medicine on core target genes. It realizes the evaluation of traditional Chinese medicine activity by replacing experience judgment with objective data, and verifies the objectivity and reliability of the algorithm of the present invention.

[0083] See Figure 2 When feeding APOE with different medications - After mouse induction, the degree of atherosclerosis in the aorta was detected by gross Oil Red O- staining. Figure 2 From left to right: control group, model group, Ator treatment group, Chiqilin formula group, low-dose new Chiqilin formula group, medium-dose new Chiqilin formula group, and high-dose new Chiqilin formula group.

[0084] Table 2. Statistics of Oil Red O staining area in each group of mice (mean ± SD, n=3)

[0085]

[0086] Table 2 shows that compared with the model group (18.89±2.71%), the Oil Red O staining area of ​​the aorta in each treatment group decreased to varying degrees. Among them, the staining area in the medium-dose new Chiqilin formula group (10.67±1.01%) was approximately 28.7% lower than that in the old Chiqilin formula group (14.95±0.74%), and was also superior to the positive control ATOR treatment group (13.82±0.66%). These results indicate that the modified compound formula containing Cnidium monnieri, saffron, Stephania tetrandra, and Alpinia oxyphylla, selected using the algorithm of this invention, can effectively reduce the formation of atherosclerotic plaques, demonstrating the targeted and efficient nature of the algorithm for improving traditional Chinese medicine compound formulas.

[0087] See Figure 3 To statistically analyze the Oil Red O-staining area and foam cell percentage in different groups of mice, thereby quantifying the degree of atherosclerosis, the medium-dose Xin Chi Qi Lin Fang group showed a significant reduction in both the atherosclerotic area and the proportion of foam cells compared to the original formula.

[0088] Table 3. Statistics on the proportion of foam cells in each group of mice (mean ± SD, n=3)

[0089]

[0090] As shown in Table 3, compared with the model group (53.12±6.47%), the proportion of foam cells in each treatment group decreased. The proportion of foam cells in the medium-dose new Chiqilin formula group (21.72±3.65%) was approximately 42.6% lower than that in the old Chiqilin formula group (37.82±2.36%), showing a significant difference. The high-dose new Chiqilin formula group (13.38±2.65%) showed the most significant effect. These results corroborate the Oil Red O staining results, further confirming that the algorithm of this invention, by integrating multi-omics data and traditional Chinese medicine targets, has good applicability and universality in its quantitative scoring-driven compound improvement strategy. It can accurately improve the efficacy of the original formula, providing objective and reliable technical support for the screening and optimization of traditional Chinese medicine compound formulas and the research and development of innovative traditional Chinese medicine drugs.

[0091] The above description is merely an example and illustration of the structure of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described, or use similar methods to replace them, as long as they do not deviate from the structure of the invention or exceed the scope defined in the claims, all of which should fall within the protection scope of the present invention.

Claims

1. An algorithm for improving and predicting traditional Chinese medicine compound prescriptions, characterized in that, Includes the following steps: S1. Obtain omics sequencing data from the disease group and the healthy group, and preprocess the omics sequencing data to construct a sample-site expression matrix; S2. Based on the database of Chinese medicine targets, establish the association between Chinese medicine and target genes, and establish a multidimensional mapping relationship between Chinese medicine, target genes and 5hmC sites based on epigenomic data to obtain the set of specific sites corresponding to each Chinese medicine. S3. Perform PCA dimensionality reduction on the target feature set of traditional Chinese medicine to extract the core principal components; S4. Calculate the association score between the sample and the traditional Chinese medicine based on the core principal components, and construct the sample-traditional Chinese medicine score matrix; S5. Conduct a significance test on the scores of the healthy group and the disease group, quantify the regulatory activity of traditional Chinese medicine by scoring the difference in activity, and screen out traditional Chinese medicines that can be used for compound improvement.

2. The algorithm according to claim 1, characterized in that, The preprocessing of the omics sequencing data includes: filtering low-quality sequencing sites, cleaning and standardizing the retained sites to construct a standardized expression matrix, and using the DEseq2 package in R language to perform variance stabilization transformation on the standardized expression matrix to eliminate interference caused by differences in sequencing depth.

3. The algorithm according to claim 1, characterized in that, The TCM target database is the HIT2 database and / or the TCMSP database. The construction of the multidimensional mapping relationship includes: retrieving the target genes of each TCM to be evaluated, forming a basic association table of TCM-target genes, matching the 5hmC sites of the corresponding target genes, and integrating them to obtain a standardized three-dimensional association matrix of TCM-target genes-5hmC sites.

4. The algorithm according to claim 1, characterized in that, The PCA dimensionality reduction of the target feature set of traditional Chinese medicine and the extraction of core principal components include: constructing a two-dimensional feature matrix of traditional Chinese medicine-feature or a three-dimensional feature matrix of sample-traditional Chinese medicine-feature based on the association data of traditional Chinese medicine-target gene or the three-dimensional association matrix of traditional Chinese medicine-target gene-5hmC site; performing PCA dimensionality reduction calculation on the feature matrix and screening core principal components with a cumulative variance contribution rate greater than or equal to a preset value.

5. The algorithm according to claim 4, characterized in that, The core principal component is a sample-TCM-TCM-specific 5hmC site matrix constructed for each TCM based on its specific 5hmC site set, and the first principal component PCA1 is extracted after PCA dimensionality reduction.

6. The algorithm according to claim 5, characterized in that, The first principal component PCA1 loading value carries the largest amount of information variation in the feature matrix, and is used to characterize the specific target and / or 5hmC site of traditional Chinese medicine regulation.

7. The algorithm according to claim 1, characterized in that, The process of calculating the association score between the sample and the traditional Chinese medicine based on the core principal components and constructing the sample-traditional Chinese medicine score matrix includes: for each sample, obtaining the score of the traditional Chinese medicine in the corresponding sample by weighted summation of the expression level of the traditional Chinese medicine-specific 5hmC site and the PCA1 loading value.

8. The algorithm according to claim 1, characterized in that, Quantifying the regulatory activity of traditional Chinese medicine through activity difference scoring includes: testing the significance of the differences in traditional Chinese medicine scores between the disease group and the healthy group; using an independent two-sample t-test to calculate the P-value of the difference in traditional Chinese medicine scores between the healthy group and the disease group; and converting the obtained P-value into activity difference scores.

9. The algorithm according to claim 1, characterized in that, The significance test for the difference between the scores of the healthy group and the disease group includes at least one of the following: t-test and Wilcoxon rank-sum test.

10. The algorithm according to claim 2, characterized in that, The low-quality sites filtered for sequencing include sites with a sequencing quality value Q < 30.

Images