Network pharmacology analysis method and model construction method of traditional Chinese medicine based on multi-scale information entropy

Abstract

The present invention provides a network pharmacology analysis method and model construction method for Chinese patent medicine based on multi-scale information entropy, and relates to the technical field of bioinformatics. It includes data collection and preprocessing, and calculating the characteristic information entropy of each scale for each data, screening out the top N ranked features of each scale, calculating the multi-scale information entropy of each component of the Chinese patent medicine, sorting the multi-scale characteristic information entropy, obtaining the top M ranked features, and constructing a drug-target-function network. It solves the technical problem that the existing technology uses traditional network analysis technology and is difficult to accurately predict the target and effect of the drug.

Description

Technical Field

[0001] The present invention belongs to the technical field of bioinformatics, and specifically relates to a network pharmacology analysis method and a model construction method for traditional Chinese medicines based on multi-scale information entropy. Background Art

[0002] As an important part of traditional Chinese medicine, Chinese patent medicine has gained more and more attention in modern society for its unique treatment concepts and methods. Chinese patent medicine has the characteristics of multiple components, multiple targets and overall regulation, which makes it show unique advantages in the treatment of complex diseases. Different from the single-target effect of Western medicine, Chinese patent medicine can comprehensively regulate multiple links of the disease through the synergistic effect of multiple components, thereby achieving the purpose of treating the disease.

[0003] Network pharmacology is an emerging research method that reveals the mechanism of action and biological effects of drugs by constructing an interaction network between drug components and biomolecules. This method is of great significance in the study of traditional Chinese medicines because it can help researchers understand the complex mode of action of traditional Chinese medicines and how they regulate human functions at the molecular level.

[0004] For example, through network pharmacology analysis, researchers found that various components in Danshen injection can protect the cardiovascular system through different signaling pathways. Using this method, researchers can more quickly screen out drug components with potential therapeutic value, or optimize existing drug formulations to improve their efficacy and safety.

[0005] For example, the Chinese invention patent application with publication number CN117612635A and titled "A method for analyzing the core active ingredients and mechanism of action of traditional Chinese medicine for strengthening spleen and removing dampness based on network pharmacology" specifically discloses the screening of the main active targets of traditional Chinese medicine for strengthening spleen and removing dampness based on network pharmacology, and the analysis of the main biological processes and pathways involved in the treatment of traditional Chinese medicine for strengthening spleen and removing dampness. The specific steps include S1: screening traditional Chinese medicine ingredients for strengthening spleen and removing dampness; S2: obtaining active ingredient targets; S3: constructing a network of active ingredients-related targets; S4: constructing a protein interaction network; S5: GO and KEGG pathway enrichment analysis; S6: screening Hubgenes in the network of related targets of traditional Chinese medicine for strengthening spleen and removing dampness; S7: determining the core active ingredients of traditional Chinese medicine for strengthening spleen and removing dampness. The network pharmacology method is applied, and combined with bioinformatics technology and computer network technology, to reveal the core active ingredients and mechanism of action of traditional Chinese medicine for strengthening spleen and removing dampness.

[0006] However, existing technologies have limitations in describing the complexity of drug molecules and predicting biological activities in the study of network pharmacology of traditional Chinese medicines. First, the ingredients of traditional Chinese medicines are complex, often containing dozens or even hundreds of compounds, and the interaction relationship between these ingredients is difficult to fully analyze using traditional experimental methods. Secondly, the mechanism of action of traditional Chinese medicines involves multiple biological molecules and signal pathways. Existing research methods can often only capture partial information and it is difficult to fully reveal its action network. In addition, existing computational methods are inefficient in processing large-scale data, and the accuracy in predicting drug biological activity needs to be improved.

[0007] For example, in one study, researchers tried to use network pharmacology methods to analyze the anti-tumor effects of a traditional Chinese medicine. However, due to the large number of drug ingredients and complex mechanism of action, traditional network analysis technology could not accurately predict the drug's targets and effects.

[0008] For example, the Chinese invention patent application with publication number CN113611372A and titled “A method and system for fitting prediction of the efficacy of effective ingredients in traditional Chinese medicine prescriptions” specifically discloses the preprocessing of the efficacy grading screening based on each monomer component of the traditional Chinese medicine prescription, which is mainly based on the data preprocessing of the total table of the traditional Chinese medicine prescription and its corresponding monomer component sub-table, and then determines the fitting score based on the preprocessed data, including the comprehensive indicator calculation method of the predictive entropy weight method, and determines the efficacy fitting score of each monomer component of the traditional Chinese medicine prescription through a comprehensive analysis of the efficacy fitting model prediction.

[0009] Previous studies have used large amounts of labeled data and only considered the structure or sequence information of drugs. Traditional structure- or sequence-based machine learning models rely heavily on feature screening and lack generalization.

[0010] Therefore, it is necessary to develop a new computational approach that should be able to better describe the complexity of drug molecules and have higher accuracy in the prediction of bioactivity. Summary of the invention

[0011] Therefore, the technical problem to be solved by the present invention is to provide a network pharmacology analysis method and model construction method for Chinese patent medicine based on multi-scale information entropy. The technical problems that the traditional machine learning models based on structure or sequence in the prior art heavily rely on feature screening and lack generalization are solved, and the complexity of drug molecules is difficult to analyze, and the accuracy of biological activity prediction needs to be improved.

[0012] The present invention provides a technical solution: a network pharmacology analysis method of Chinese patent medicine based on multi-scale information entropy, the analysis method comprising the following steps:

[0013] S1. Data collection, including information on components of Chinese patent medicines, small molecule compounds, target genes and interaction information, gene functions and metabolic pathways;

[0014] S2, preprocessing the collected data;

[0015] S3, calculating the feature information entropy of each scale for each data, setting a first threshold N, and screening out the top N ranked features of each scale;

[0016] S4, calculating the multi-scale information entropy of each component of the Chinese patent medicine, setting a second threshold M, sorting the multi-scale feature information entropy, and obtaining the top M sorted features;

[0017] Construct drug-target-function network.

[0018] Preferably, the characteristic information entropy calculation process of each scale in step S3 is:

[0019] ;

[0020] in, Indicates j Information entropy of scale, Indicates j The scale i The probability of an event occurring, n is the total number of events.

[0021] Preferably, the data information defined by the scale includes the scale of molecular structure, the scale of drug efficacy target, the scale of drug-target gene interaction, the scale of gene function and metabolic pathway.

[0022] Preferably, the data preprocessing in step S2 includes cleaning, formatting and normalizing the data.

[0023] Preferably, the multi-scale information entropy of each component of the Chinese patent medicine in step S4 is calculated by:

[0024] ;

[0025] in, represents the information entropy of m scales, represents the weight of the jth scale, and m represents the total number of scales.

[0026] Preferably, the analysis method further comprises the steps of:

[0027] S5. Use the multi-scale information entropy of each component of the Chinese patent medicine calculated in S4 as a feature to perform machine deep learning.

[0028] Preferably, the multi-scale information entropy of each component of the Chinese patent medicine includes structural entropy, target entropy, interaction entropy and functional entropy.

[0029] Preferably, the Chinese patent medicine includes Compound Danshen Tablets.

[0030] The present invention provides another technical solution: a method for constructing a model of network pharmacology of Chinese patent medicines based on multi-scale information entropy, characterized in that the model construction method comprises the following steps:

[0031] S1. Data collection, including information on components of Chinese patent medicines, small molecule compounds, target genes and interaction information, gene functions and metabolic pathways;

[0032] S2, preprocessing the collected data;

[0033] S3. Calculate the feature information entropy of each scale for each data, and select the top N ranked features of each scale. The feature information entropy calculation process of each scale is as follows:

[0034] ;

[0035] in, Indicates j Information entropy of scale, Indicates j The scale i The probability of an event occurring, n represents the total number of events;

[0036] S4. Calculate the multi-scale information entropy of each component of the Chinese patent medicine, sort the multi-scale feature information entropy, and obtain the features of the top M sorts. The calculation process is:

[0037] ;

[0038] in, represents the information entropy of m scales, represents the weight of the jth scale, and m represents the total number of scales;

[0039] S5. The multi-scale information entropy of each component of the Chinese patent medicine calculated in S4 is used as a feature to build a deep learning model, and the up-regulation, down-regulation or no effect of each target is used as the target variable; the Chinese patent medicine and the target are randomly combined to obtain the sample data of each combination, and the up-regulated and down-regulated targets are selected as positive samples in the Chinese patent medicine component target database, and other unknown labels are used as negative samples.

[0040] Preferably, the input data of the model are information entropies of various scales calculated based on the components of the Chinese patent medicine, including structural entropy, target entropy, interaction entropy, and functional entropy; as well as data on the differential gene expression levels of each sample after the intervention of the Chinese patent medicine; the output data of the model is whether the Chinese patent medicine has an effect on the target.

[0041] Beneficial effects:

[0042] The network pharmacology analysis method of Chinese patent medicine based on multi-scale information entropy provided by the present invention includes data collection and preprocessing, and calculating the characteristic information entropy of each scale for each data, screening out the features of the top N rankings at each scale, calculating the multi-scale information entropy of each component of the Chinese patent medicine, sorting the multi-scale characteristic information entropy, obtaining the features of the top M rankings, and constructing a drug-target-function network. It overcomes the problem that the prior art uses traditional network analysis technology, which makes it difficult to accurately predict the target and effect of the drug. The method of the present invention collects and preprocesses multi-dimensional data of Chinese patent medicines, and then calculates multi-scale information entropy, quantifies the diversity and action path of drug components, and then analyzes the interaction and mechanism of action of the components of Chinese patent medicines based on deep learning methods.

[0043] The model construction method of the network pharmacology of Chinese patent medicines with multi-scale information entropy provided by the present invention more accurately describes the complexity of Chinese patent medicine ingredients and predicts their biological activity by developing a calculation method of a deep learning model based on multi-scale information entropy. The interaction between Chinese patent medicine compound and target molecules is often highly nonlinear. The deep learning model can capture these complex nonlinear relationships, and can learn the relationship between the complex characteristics of the Chinese patent medicine compound and its targeted molecular mechanism of action, which helps to more accurately predict the effect of Chinese patent medicine. The model constructed by the method of the present invention can not only improve research efficiency and reduce experimental costs, but also provide a more scientific theoretical basis for the research and development of Chinese patent medicines, and promote the development of modernization of traditional Chinese medicine. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to make the contents of the present invention more clearly understood, the present invention is further described in detail below based on specific embodiments of the present invention in conjunction with the accompanying drawings.

[0045] Figure 1 This is a schematic diagram of the functional pathway results involved in the compound danshen tablets in Example 2 of the present invention. DETAILED DESCRIPTION

[0046] The present invention will be described in detail below with reference to the accompanying drawings and in combination with embodiments. The principles and features of the present invention are described below in combination with the accompanying drawings. It should be noted that the embodiments in this application and the features in the embodiments can be combined with each other without conflict. The embodiments are only used to explain the present invention and are not used to limit the scope of the present invention.

[0047] Example 1

[0048] This embodiment is a network pharmacology analysis method for Chinese patent medicine based on multi-scale information entropy, comprising the following steps:

[0049] S1. Data collection, the data includes component information, small molecule compounds, target genes and interaction information, gene function and metabolic pathway information of Chinese patent medicine. For example, the components of Chinese patent medicine are searched in the TCMSP database of the Chinese medicine system pharmacology database and analysis platform to obtain the compounds of each component. The Smile structure of the compound is obtained in the compound structure database Pubchem to obtain the structural information of the small molecule compound. The information of the target gene corresponding to the small molecule compound is collected in the TargetNet database. Gene interaction information is obtained through the StringDB database. From the GO and KEGG databases, it is obtained which GO gene functions and KEGG metabolic pathways each gene is involved in.

[0050] S2. Preprocess the collected data, including cleaning, formatting and normalizing the data, that is, filtering the collected data, SMILE formatting of the structural information, feature input for deep learning, and calculating the probability of event occurrence based on correlation.

[0051] S3. Calculate the feature information entropy of each scale for each data, set the first threshold N, and filter out the top N ranked features of each scale, including:

[0052] S3.1. Define the data information of each scale of the components of Chinese patent medicines. Each scale represents a different scale. The data information defined by the scale includes the molecular structure scale, the drug efficacy target scale, the drug target gene interaction scale, the gene function and the metabolic pathway scale.

[0053] S3.2, the calculation process of feature information entropy at each scale is:

[0054] ;

[0055] in, Indicates j Information entropy of scale, Indicates j The scale i The probability of an event occurring, n is the total number of events.

[0056] Information entropy measures the amount of information. In this method, the smaller the information entropy, the less random information the scale feature has, thus providing more information. The information entropy of each scale feature can be sorted, and the TopN information entropy features can be screened at each scale to reduce the impact of data noise.

[0057] S4, calculate the multi-scale information entropy of the components of the Chinese patent medicine, that is, the joint information entropy of multiple scales; set the second threshold M, sort the multi-scale feature information entropy, and obtain the features of the top M sorts. The calculation process is:

[0058] ;

[0059] in, represents the information entropy of m scales, represents the weight of the jth scale, and m represents the total number of scales.

[0060] Multi-scale feature information entropy sorting is performed, and the multi-scale information entropy features of TopM of Chinese patent medicine components are selected to reduce the impact of data noise.

[0061] Furthermore, it also includes:

[0062] S5. The multi-scale information entropy of each component of the Chinese patent medicine calculated in S4 is used as a feature to perform machine deep learning; the information entropy of each scale calculated is used as a feature, which reflects the complexity and uncertainty of the system, and the up-regulation, down-regulation or no effect of the Chinese patent medicine on each gene target is used as the target variable. The Chinese patent medicine and the gene target are randomly combined to obtain the sample data of each combination, and the up-regulated and down-regulated gene targets are selected as positive samples in the gene target database corresponding to the Chinese patent medicine ingredients, and other unknown labels are used as negative samples.

[0063] Among them, the information entropy of each scale of each component of Chinese patent medicine includes the structural entropy, target entropy, interaction entropy and functional entropy of each component.

[0064] Construct a drug-target-function network, including the TopM Chinese patent medicine components, the information entropy characteristics of the TopN screened at each scale, and the predicted positive targets, to reveal the complex mechanism of action of the Chinese medicine compound from a holistic and systematic perspective.

[0065] Example 2

[0066] This embodiment is a network pharmacology analysis method of compound Danshen tablets based on multi-scale information entropy, wherein the compound Danshen tablets are used to treat cardiovascular diseases. The analysis method based on Example 1 comprises the following steps:

[0067] S1. Data collection: The component information of compound Danshen tablets includes Danshen, Panax notoginseng, and Borneol. The small molecule compounds of each component were obtained by searching the database of TCMSP, a Chinese medicine system pharmacology database and analysis platform:

[0068] The compounds of Salvia miltiorrhiza include Sitosterol, Î’-Sitosterol, Stigmasterol, Oleanolic Acid, Alexandrin, Daucosterol, Eleutheroside A, Succinic Acid, Alexandrin, Daucosterol, Caproic Acid, Eleutheroside A, Sitogluside, Strumaroside, Î’-Sitosterol-Î’-D-Glucoside, 3,4-Dihydroxybenzoic Acid, Protocatechuic Acid, Isoimperatorin, Methylene Tanshinquinone, Protocatechuic Aldehyde, 5,7,4'-Trihydroxy-6,8,3'-Trimethoxyflavone, Labiatenic Acid, Rosmarinic Acid, Ursolic Acid, 3-O-Acetyloleanolic Acid, Dihydrotanshinone I, Dihydroisotanshinone I, 1,5-Dihydroxy-3-Methylanthraquinone、Tanshinlactone、Salvianolic Acid B、Salvianolic Acid D、Salvianolic Acid E、DanshexinkumD、Danshexinkum A、Methyl Tanshinonate、Tanshinaldehyde、Isocryptotanshinone、Paramiltioic Acid、Ferruginol、Sugiol、Tigogenin、EthylLithospermate、Salvinone、Neocryptotanshinone、Isotanshinoneii B、Prgewaquinone A、Baicalin、Cryptotanshinone、Danshenol A、DanshenolB、Danshenspiroketallactone、Danshensu、Danshensuan B、Danshenxinkun A、Danshenxinkun B、DanshenxinkunC、Danshenxinkun D、Dehydromiltirone、Δ1-Dehydrotanshinone、Dihydroisotanshinone I、Dihydrotanshinone I、1,2-Dihydrotanshiquinone、Epidanshenspiroketallactone、Ferruginol、Hesperetic Acid、3-Î’-Hydroxymethylenetanshiquinone、3Α-Hydroxytanshinone Iia、3Î’-Hydroxytanshinone Iia、Isocryptotanshinone、2-Isopropyl-8-Methylphenanthrene-3,4-Dione(R0-090680)、Isotanshinone I、IsotanshinoneIia、1-Keto-Isocryptotanshinone、Lithospermate B、Magnesium Lithospermate B、MethylTanshinonate、Miltionone I、MiltiononeIi、Miltipolone、Miltirone、MonomethylLithospermate、Neocryptotanshinone、Neocryptotanshinone Ii、Nortanshinone、OleoylDanshenxinkun A、Oleoyl Neocryptotanshinone、Przewaquinone B、RosmarinicAcidMethyl Ester、Salvianolic Acid A、Salvianolic Acid B、Salvianolic Acid C、Salvianolic Acid D、Salvianolic Acid E、SalvianolicAcid G、Salvilenone、Salvinone、Salviol、Tanshindiol A、Tanshindiol B、Tanshindiol C、Tanshinlactone、Tanshinol A、TanshinolB、Tanshinone I、Tanshinone Iia、Tanshinone Iib、TanshinoneVi、(25R)-5Α-Spirostan-3Î’-Ol,Tigogenin。,

[0069] Also known as:Î'-Guriunene、Sitosterol,Î'-Sitosterol、(1S,2S)-2-Isopropenyl-4-Isopropylidene-1-Methyl- 1-Vinylcyclohexane,Γ-Elemene、Guerketol,Quercetin,Quercetin,Sophoretin,Meletine,Xanthaurine、Ethyl Palmitate、(1R,9S)-4,11,11-Trimethyl-8-Methylenebicyclo[7.2.0]Undec-4-Ene,Caryophellene,Î'-Caryophyllene、Stigmasterol、Cetylic Acid,Hexadecanoic Acid,Pal Acid、Nonadecane、Alexandrin,Daucosterol,CaproicAcid,Eleutheroside A,Cytogluside,Strumaroside,Î'-Sitosterol-Î'-D-Glucoside、Hexadecane、Beta-Elemene,Î'-Elemene、'-Guaiene、Δ-Guaiene、Enanthic Acid、Nonanoicthy Acid、Nonanoicthy Acid Hexadecanate,Methyl Palmitate、Docosane、Heneicosane、HeneicosanoicAcid、Heptadecane、Pentade cane、Eicosane、Copaene、Cuparene、Cyclo-(Ala-Pro)、Cyclo-(Ala-Val)、Δ-Cadinol,Δ-Cadinene、2,6-Ditertbutyl-4-Methyl Phenol、Ginsenoside Rb1、Ginsenoside Rc、Ginsenoside Rg1、Ginsenoside-Rh1、Α-Muurolene、NotoginsenosideR1、Notoginsenoside R4、Panaxydol、Panaxytriol、Tetradecane、Acetophenone、L-Α-Amino-Î’-Oxalylaminopropionic Acid、Cyclododecanone、Cyperene、Dicapryl Phthalate、Diisocapryl Phthalate、2,8-Dimethyl-5-Acetyl-Bicyclo[5,3,0]Decadiene-1,8、Α,Α-Dimethylbenzene Methanol、DitertbutylPhthalate、3-Ene-Nonanone-2、3Î’-O-Î’-D-Glucopyranosyl-20-O-[Α-L-Arabinopyranosyl(1→2)-Î’-D-Glucopyranosyl]3Î’,12Î’,20(S)-Trihydroxydammar-24-Ene、3-O-[Î’-D-Glucopyranosyl(1→6)-Î’-D-Glucopyranosyl]-20-O-Î’-D-Glucopyranosyl-3Î’,12Î’,20(S)-Trihydroxydammar-24-Ene、6-O-Î’-D-Glucopyranosyl-20-O-Î’-D-Glucopyranosyl-3Î’,6Α,12Î’,20(S),25-Pentahydroxydammar-23-Ene、1-Methyl-4-Isoallyl-Cyclohexane、Notoginsenoside A、Notoginsenoside B、Notoginsenoside C、Notoginsenoside D、NotoginsenosideG、Notoginsenoside H、Notoginsenoside I、Notoginsenoside J、Notoginsenoside K、Notoginsenoside L、NotoginsenosideM、Notoginsenoside N、Notoginsenoside R2、Notoginsenoside R3、Notoginsenoside R6、(S)-(+)-Falcarinol,Panaxynol、Quinquenoside I、QuinquenosideIi、Quinquenoside Iii、Quinquenoside Iv、Quinquenoside V、Sanchinoside B1、1,1,5,5-Tetramethyl-4-Methano-2,3,4,6,7,10-Hexahydronaphthalene、Tridecene、2,2,2-Triethoxyl-Ethanol、20-O-Î’-D-Xylopyranosyl(1→6)-Î’-D-Glucopyranosyl-20(S)-Protopanaxadiol。,

[0070] Borneol compounds include Alpha-Humulene, Humulene, Î'-Humulene, 1,7,7-Trimethylbicyclo[2.2.1]Heptan-2-Ol, Borneol, Camphor, (1R,9S)-4,11,11-Trime thyl-8-Methylenebicyclo[7.2.0]Undec-4-Ene,Caryophellene,Î'-Caryophyllene,OleanolicAcid,Dl-Isoborneol,AlphitolicAcid,Asiatic Acid, Borneol, Borneol, Camphor, Î'-Caryophyllene, Dipterocarpol, Dryobalanone, Beta-Elemene, Î'-Elemene, Erythrodiol, D-Isoborneol, (1R,2R,4R)-1,7,7-Trimethylbicyclo[2.2.1]Heptan-2-Ol,Isoborneol,L-Isoborneol.

[0071] Collect information on target genes corresponding to small molecule compounds in the TargetNet database:

[0072] The targets of Salvia miltiorrhiza include HSD17B1; IL1B; HSD17B2; COX6A2; CES1; BGLAP; ASPH; MT-CO1; NFKBIA; MIF; FOS; MAPK1; P4HA1; HSD17B6; MT-CO3; P4HA2; ESRRA; PRKCA; JAK1; ESRRB; PRKCB; CSNK2A1; PLOD1; SUCNR1; SF3B3; TMLHE; GABRD; TP53; SLC13A1; SDHA; PLOD3; PPARA; GABRE; ATP6; EGFR; SDHB; SLC13A2; ESRRG; SQLE; SDHC; TYR; SLC13A3; SDHD; AKR1D1; TOP2A; GLRA3; SULT2B1; PPARG; CASP3; CLEC4E; DHFRL1; PIM1; ITGAL; HSP90AA1; RAC1; SRD5A2; RUVBL2; NQO1; NQO2; GABRP; NR1I2; GABRQ; NR1I3; LSS; AKT1; CA12; CHRNA4; BBOX1; HBB; HBA1; TNF; CA14; COX7A1; OXCT1; OXCT2; DGKA; PROS1; GRIN1; PTGIR; PTGIS; LIP3; PIK3R1; PTGS1; GABRA1; HIBCH; GABRA2; COX5A; GABRA3; COX5B; GABRA4; F2; GABRA5; PROC; GABRA6; COMT; GRIN3A; ABCC5; F7; GRIN3B; NCOA1; COX7B; GPBAR1; DNMT1; NCOA2; ESR1; SLCO1B3; COX7C; F9; EIF3F; ESR2; EFTUD1; ALDH5A1; PPP1CC; ANXA1; CA5A; CA5B; FECH; CA1; NR3C1; GABRG1; CA2; PLA2G1B; NR3C2; GABRG2; CA3; GABRG3; CA4; ITGB2; VDR; STK17B; CA6; CA7; PIK3CA; GSTP1; PROZ; CSNK2B; CA9; COX6B1; SMO; ALAD; PIK3CG; P2RY12; LPL; SEC14L2; ACTB; SEC14L3; CYP1B1; SEC14L4; AKR1C1; NPPB; SULT2A1; AKR1C2; CDK6; PPP2CA; PPP2CB; UBA1; P3H1; F10; P3H2; P3H3; NR1H4; SUCLA2; HSD11B1; HSD11B2; GPER1; NFKB1;NFKB2ATP5BPYGMPGRCYP19A1PLK1PRKACAVKORC1FABP6HSD3B1SIGMAR1COX4I1GGCXACO2 ATP5A1ALOX5HCKGABRB1GRIN2AGABRB2CBR1HOXA10GRIN2BGABRB3SUCLG1GRIN2CSYKSU CLG2COX6CGRIN2DBECN1ATP5C1PRLRSHBGRBP3ADH1CHSPA2COX8ALCN2PTK2BYWHAEVKOR C1L1,RORA,IGHG1,ATP1A1,IGHG2,ATP1A2,ATP1A3,STAT3,UGT3A1,HMGCR,SLC25A10,AHR,CEBPB.

[0073] The targets of Panax notoginseng include: PLA2G2D; CALM1; HSD17B1; PLA2G2E; SCN2A; IL1B; SERPINA1; SCN2B; PRKCA; ESRRA; SCN4A; PRKCB; JAK1; ESRRB; SCN4B; CSNK2A1; SF3B3; GABRD; ECI2; PPARA; GABRE; SQLE; PPARD; TOP2A; ATP2A1; SULT2B1; GLRA3; PPARG; DHFRL1; PIM1; CLEC4E; CASP3; HSP90AA1; RUVBL2; ELOVL4; SCN8A; NQO2; GABRP; NR1I2; AKT1; GABRQ; NR1I3; LSS; CA12; CA14; TNF; SLC8A1; OGDH; FURIN; GRIN1; ACOT13; RAC1; LIP3; PTGS1; GABRA1; SCN10A; PTGS2; HIBCH; GABRA2; GABRA3; GABRA4; GABRA5; COMT; GABRA6; ARF1; ADH5; GRIN3A; NCOA1; GRIN3B; LALBA; DNMT1; ESR1; NCOA2; CYP2C8; EIF3F; ESR2; SLCO1B3; PPT1; HAO1; ARF6; ALDH5A1; PAEP; PPP1CC; ANXA1; CA5A; CA5B; CA1; CA2; NR3C1; GABRG1; PLA2G1B; ABAT; CA3; GABRG2; NR3C2; CNR1; CA4; GABRG3; ACADSB; CNR2; SOAT1; JUN; STK17B; VDR; SCN1A; SOAT2; CA6; HA; SCN1B; CA7; CSNK2B; CA9; ACSL3; GUCA1A; ACSL4; TRAPPC3; SCN3A; HDAC2; SCN3B; PIK3CG; SREBF1; LPL; TRPV1; SEC14L2; ACTB; S100B; HNF4A; CYP1B1; LTB4R; SCN5A; HDAC9; PKIA; RCVRN; AKR1C1; CDK6; SULT2A1; AKR1C2; IL6; PMP2; LTF; HNF4G; PPP3R1; UBA1; SCN7A; PVR; NR1H4; RHO; HSD11B1; PPA_RS05235; GPER1; HSD11B2; LY96; NFKB1; SCN9A; NFKB2; ATP5B; PGR; CYP19A1; FADS1; PRKACA; FADS2; ABO; TM1468; SIGMAR1;HSD3B1; GLTP; ATP5A1; HCK; MTTP; GABRB1; SCN11A; CBR1; GRIN2A; GABRB2; GRIN2B; GABRB3; PPP3CA; SYK; GRIN2C; GRIN2D; ATP5C1; RARG; SHBG; C8G; HSPA2; RXRA; RXRB; INS; LCN2; PTK2B; KANSL3; RORA; YWHAE; IGHG1; ATP1A1; FFAR1; ATP1A2; JUN; RXRG; TLR4; FKBP1A; UGT3A1; AR; ATP1A3; ACOX1; GM2A; VLDLR; AHR; PLA2G2A; CEBPB; LTB4R2; ABL1 etc.;

[0074] Borneol targets include: HSD17B1; IL1B; CES1; COX6A2; MT-CO1; MT-CO2; MT-CO3; GABRD; PPARA; GABRE; ESRRG; AKR1D1; SULT2B1; CLEC4E; SRD5A1; GABRP; NR1I2; GABRQ; NR1I3; LSS; TNF; COX7A1; CACNA1C; CACNA1D; CACNA1F; GRIN1; NR0B 1; LIP3; PTGS1; GABRA1; GABRA2; COX5A; GABRA3; COX5B; GABRA4; GABRA5; GABRA6; GRIN3A; CACNA1S; GRIN3B; N COA1; COX7B; NCOA2; ESR1; GPBAR1; SLCO1B3; ESR2; COX7C; EFTUD1; PPP1CC; ANXA1; FECH; GABRG1; NR3C1; PLA2 G1B; GABRG2; NR3C2; GABRG3; VDR; TRPA1; GSTP1; COX6B1; HDAC2; TRPV1; TRPV3; AKR1C1; IL6; SULT2A1; NPPB; A KR1C2;CACNB1;CACNB2;CACNB3;HSD17B11;CACNB4;NR1H4;HSD11B1;HSD11B2;NFKB1;NFKB2;CYP27B1;PGR;F ABP6; HSD3B1; SIGMAR1; COX4I1; ALOX5; GABRB1; MED1; GABRB2; GRIN2A; HOXA10; GABRB3; GRIN2B; GRIN2C; NOS 2; GRIN2D; COX6C; SHBG; PRLR; OPRK1; TRPM8; ADH1C; RXRA; COX8A; RORA; ATP1A1; IGHG2; ATP1A2; ATP1A3; AR, etc.

[0075] Gene interaction information was obtained through the StringDB database. The GO and KEGG databases were used to obtain the GO gene functions and KEGG metabolic pathways in which each gene was involved. Gene functions include molecular functions, cell components, biological processes and other functional information in the GO database. Metabolic pathways include relevant pathway information such as immunity and metabolism in the KEGG database, as shown in Table 1 below.

[0076] Table 1 Genes obtained from GO and KEGG databases

[0077]

[0078] Function information corresponding to Table 1:

[0079] MAPK1, also known as mitogen-activated protein kinase 1, is a member of the MAP kinase family. It acts as an integration point for multiple biochemical signals and is involved in a variety of cellular processes such as proliferation, differentiation, transcriptional regulation, and development. Activation of MAPK1 requires phosphorylation by upstream kinases. Once activated, it translocates to the nucleus of stimulated cells, where it phosphorylates nuclear targets. In addition, MAPK1 has been identified as a moonlighting protein capable of performing mechanistically distinct functions. Two alternatively spliced ​​transcript variants encoding the same protein, but with different UTRs, have been reported.

[0080] AKT1 is a gene encoding a protein kinase and a member of the AKT serine-threonine protein kinase family. The family includes three highly similar proteins, AKT, α, β, and γ, which regulate a variety of cellular functions, including cell proliferation, survival, metabolism, and angiogenesis. AKT protein is phosphorylated by phosphoinositide 3-kinase (PI3K) and recruited to the cell membrane by phosphatidylinositol 3,4,5-triphosphate (PIP3). The PI3K / AKT signaling pathway is essential for tumor cell survival, and mutations in the AKT1 gene are associated with multiple types of cancer and excessive tissue growth. AKT protein is also involved in the mammalian target of rapamycin (mTOR) signaling pathway, which is dysregulated in many cancers.

[0081] The FOS gene family includes four members, FOS, FOSB, FOSL1, and FOSL2, which encode leucine zipper proteins that can bind to JUN family proteins to form the transcription factor complex AP-1. FOS proteins are involved in regulating cell proliferation, differentiation, and transformation, and in some cases, they are associated with apoptotic cell death. The FOS gene is a proto-oncogene, which means that it has the potential to cause cancer if mutated or overexpressed. Overall, the FOS gene family plays a vital role in various cellular processes and is a topic of ongoing research in the field of molecular biology.

[0082] The TP53 gene encodes a tumor suppressor protein that regulates the expression of target genes in response to various cellular stresses, inducing cell cycle arrest, apoptosis, senescence, DNA repair, or metabolic changes. The protein contains transcriptional activation, DNA binding, and oligomerization domains. Mutations in this gene have been associated with various human cancers, including hereditary cancers such as Li-Fraumeni syndrome. Alternative splicing and the use of alternative promoters result in multiple transcript variants and isoforms. Other isoforms have also been shown to arise from the use of alternative translation start codons from the same transcript variant.

[0083] STAT3 is a gene encoding a protein of the STAT family. This protein is activated by phosphorylation under the action of various cytokines and growth factors, forming homo- or heterodimers and transported to the nucleus, where it acts as a transcriptional activator. STAT3 plays a key role in many cellular processes such as cell growth and apoptosis, and mediates the expression of multiple genes under cell stimulation. This protein is regulated by the small GTPase Rac1 and inhibited by the PIAS3 protein. Mutations in this gene are associated with infantile multisystem autoimmune disease and hyperimmunoglobulin E syndrome, and play a role in regulating the host response to viral and bacterial infections.

[0084] IL6, or interleukin 6, is a gene that produces a cytokine that is involved in inflammation and the maturation of B cells. The protein produced by this gene can cause fever in people with autoimmune diseases or infections and is primarily produced at sites of acute and chronic inflammation. IL6 has been implicated in a variety of inflammation-related disease states, including diabetes and systemic juvenile rheumatoid arthritis. Elevated levels of this protein have been found in viral infections, including COVID-19.

[0085] The JUN gene is a proto-oncogene that encodes a protein that interacts with specific target DNA sequences to regulate gene expression. It is highly similar to the viral protein of avian sarcoma virus 17 and is involved in translocations and deletions in human malignancies. The gene is located on chromosome 1p32-p31 and is intron-free. The putative transforming capacity suggests that it may play a role in the development of cancer.

[0086] Epidermal growth factor receptor (EGFR) is a transmembrane glycoprotein that belongs to the protein kinase superfamily. It acts as a receptor for members of the epidermal growth factor family and is involved in cell proliferation. Mutations in this gene are associated with lung cancer. EGFR is also a component of the cytokine storm that contributes to the severe form of COVID-19 caused by SARS-CoV-2. When EGFR binds to EGFR, it induces receptor dimerization and tyrosine autophosphorylation, leading to cell proliferation.

[0087] CASP1, also known as caspase 1, is a gene that encodes a protein of the caspase family of cysteine-aspartate proteases. Caspases are involved in the process of apoptosis, they exist as inactive proenzymes that are activated through a proteolytic process. CASP1 was identified for its ability to activate interleukin-1, a cytokine involved in inflammation, septic shock, and wound healing. The gene has been shown to induce apoptosis and may play a role at different stages of development. Alternative splicing of the gene produces different isoforms. Studies in mice suggest that similar genes may be involved in the pathogenesis of Huntington's disease.

[0088] RAC1 is a gene encoding a GTPase that belongs to the RAS superfamily of small GTP-binding proteins. This superfamily regulates various cellular events, including cell growth, cytoskeletal reorganization, and protein kinase activation. This gene has two transcript variants encoding different isoforms.

[0089] S2. Preprocess the collected data, specifically cleaning, formatting and normalizing the data, that is, filtering the collected data, SMILE formatting of the structural information, feature input for deep learning, and calculating the probability of event occurrence based on correlation;

[0090] S3, calculate the feature information entropy of each scale for each data, and select the top N ranked features of each scale;

[0091] S3.1. Define the data information of each scale of Danshen, Panax notoginseng and Borneol in the components of Compound Danshen Tablets. The data information defined in the scale includes the scale of molecular compounds, the scale of gene targets and the scale of functional pathways.

[0092] S3.2, the calculation process of feature information entropy at each scale is:

[0093] ;

[0094] in, Indicates j Information entropy of scale, Indicates j The scale i The probability of an event occurring, n is the total number of events.

[0095] For example, among the small molecule compounds {A, B, C, D}, if a small molecule compound is related to other compounds, and their mutual correlation is {(A, B), (A, C), (B, C), (B, D)}, the probability of the occurrence of small molecule compound A is 2 / 8. The vector is [0.25, 0.375, 0.25, 0.125], the information entropy vector of each small molecule compound is [0.5, 0.53, 0.5, 0.375], and the order of small molecule compounds is B, A, C, D. Similarly, the information entropy features of other scales are calculated.

[0096] The information entropy of each scale feature is sorted, and the characteristics of the TopN information entropy are screened at each scale to reduce the impact of data noise. In this embodiment, the first threshold N=10 is set, and the characteristics of the Top10 information entropy are screened at each scale as follows:

[0097] TOP10 gene targets of information entropy: MAPK1, AKT1, FOS, TP53, STAT3, IL6, JUN, EGFR, CASP3, RAC1.

[0098] The top 10 molecular compounds are: danshenspiroketallactone, cryptotanshinone, stigmasterol, ginsenoside, rh2deoxyneocryptotanshinone, epidanshenspiroketallactone, isotanshinone II, miltionone Ⅱ, mandenol and bronylacetate.

[0099] TOP10 functional pathways: response to stimulus, positive regulation of biological process, metabolic process, cellular process, negative regulation of biological process, locomotion, multicellular organismal process, developmental process, biological process involved in interspecies interaction between organisms, and immune system process.

[0100] S4. Calculate the multi-scale information entropy of the components of the compound Danshen tablets, sort the multi-scale feature information entropy, and obtain the features of the top M sorts. The calculation process is:

[0101] ;

[0102] in, represents the information entropy of m scales, represents the weight of the jth scale, and m represents the total number of scales.

[0103] After selecting the features of the TopN (N=10) of each scale based on S3, the weight vectors of each scale are set and weighted summation is performed to obtain the multi-scale information entropy of each component, and the multi-scale feature information entropy is sorted, and the features of the multi-scale information entropy of the TopM component of the compound Danshen tablet are selected to reduce the impact of data noise. In this embodiment, the components of the compound Danshen tablet are Danshen, Panax notoginseng, and Borneol, and the second threshold M=3 is set.

[0104] S5. The multi-scale information entropy of each component of Compound Danshen Tablet (Danshen, Panax notoginseng, Borneol) calculated in S4 is used as a feature to perform machine deep learning and build a deep learning model; the information entropy of each scale calculated is used as a feature, which reflects the complexity and uncertainty of the system, and the up-regulation, down-regulation or no effect of Compound Danshen Tablet on each gene target is used as the target variable. The components of Compound Danshen Tablet (Danshen, Panax notoginseng, Borneol) and gene targets (gene targets of the above TOP10 information entropy) are randomly combined to obtain sample data of each combination, and the up-regulated and down-regulated gene targets are selected as positive samples in the gene target database corresponding to the components contained in each component of Compound Danshen Tablet (i.e., the molecular compounds contained in each component) as positive samples, and other unknown labels are used as negative samples.

[0105] Among them, the information entropy of each scale of each component of Compound Danshen Tablets includes the structural entropy, target entropy, interaction entropy, and functional entropy of each component.

[0106] Construct a drug-target-function network, including the TopM Chinese patent medicine components, the information entropy characteristics of the TopN screened at various scales, and the predicted positive targets, to reveal the complex mechanism of action of the Chinese medicine compound from a holistic and systematic perspective.

[0107] In this embodiment, based on the deep learning model, the information entropy feature vectors of the compound danshen tablets are input, and the prediction results of the interaction between the compound danshen tablets and each target are output, see Table 2 below.

[0108] Table 2 Prediction results of the interaction between Fufang Danshen Tablets and various gene targets

[0109]

[0110] Example 3

[0111] This embodiment is a method for constructing a model of network pharmacology of Chinese patent medicines based on multi-scale information entropy, comprising the following steps:

[0112] S1. Data collection, the data includes component information, small molecule compounds, target genes and interaction information, gene function and metabolic pathway information of Chinese patent medicine. For example, the components of Chinese patent medicine are searched in the TCMSP database of the Chinese medicine system pharmacology database and analysis platform to obtain the compounds of each component. The Smile structure of the compound is obtained in the compound structure database Pubchem to obtain the structural information of the small molecule compound. The information of the target gene corresponding to the small molecule compound is collected in the TargetNet database. Gene interaction information is obtained through the StringDB database. From the GO and KEGG databases, it is obtained which GO gene functions and KEGG metabolic pathways each gene is involved in.

[0113] S2. Preprocess the collected data, including cleaning, formatting and normalizing the data, that is, filtering the collected data, SMILE formatting of the structural information, feature input for deep learning, and calculating the probability of event occurrence based on correlation.

[0114] S3. Calculate the feature information entropy of each scale for each data, set the first threshold N, and filter out the top N ranked features of each scale, including:

[0115] S3.1. Define the data information of each scale of the components of Chinese patent medicines. Each scale represents a different scale. The data information defined by the scale includes the molecular structure scale, the drug efficacy target scale, the drug target gene interaction scale, the gene function and the metabolic pathway scale.

[0116] S3.2, the calculation process of feature information entropy at each scale is:

[0117] ;

[0118] in, Indicates j Information entropy of scale, Indicates j The scale i The probability of an event occurring, n is the total number of events.

[0119] Information entropy measures the amount of information. In this method, the smaller the information entropy, the less random information the scale feature has, thus providing more information. The information entropy of each scale feature can be sorted, and the TopN information entropy features can be screened at each scale to reduce the impact of data noise.

[0120] S4. Calculate the multi-scale information entropy of each component of the Chinese patent medicine, that is, the joint information entropy of multiple scales; set the second threshold M, sort the multi-scale feature information entropy, and obtain the features of the top M sorts. The calculation process is:

[0121] ;

[0122] in, represents the information entropy of m scales, represents the weight of the jth scale, and m represents the total number of scales.

[0123] Multi-scale feature information entropy sorting is performed, and the multi-scale information entropy features of TopM of Chinese patent medicine components are selected to reduce the impact of data noise.

[0124] S5. The multi-scale information entropy of each component of the Chinese patent medicine calculated in S4 is used as a feature to build a deep learning model, and the up-regulation, down-regulation or no effect of each gene target is used as the target variable; the Chinese patent medicine and the target are randomly combined to obtain the sample data of each combination, and the up-regulated and down-regulated targets are selected from the Chinese patent medicine component target database as positive samples, and other unknown labels are used as negative samples.

[0125] The input data of the model are the information entropies of various scales calculated based on the components of Chinese patent medicine, including the structural entropy, target entropy, interaction entropy, and functional entropy of the components; as well as the data of the differential gene expression levels of each sample after the intervention of Chinese patent medicine; the output data of the model is whether the Chinese patent medicine has an effect on the target, see Table 3.

[0126] Table 3 Output results of whether Chinese patent medicines act on each target

[0127]

[0128] Obviously, the above embodiments are merely examples for the purpose of clear explanation, and are not intended to limit the implementation methods. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the implementation methods here. The obvious changes or modifications derived therefrom are still within the scope of protection of the invention.

Claims

1. A network pharmacology analysis method for Chinese patent medicine based on multi-scale information entropy, characterized in that: The analytical method comprises the following steps: S1. Data collection, including information on components of Chinese patent medicines, small molecule compounds, target genes and interaction information, gene functions and metabolic pathways; S2, preprocessing the collected data; S3, calculating the feature information entropy of each scale for each data, setting a first threshold N, and screening out the top N ranked features of each scale; The characteristic information entropy calculation process of each scale is as follows: ； in, Indicates j Information entropy of scale, Indicates j The scale i The probability of an event occurring, n represents the total number of events; S4, calculating the multi-scale information entropy of each component of the Chinese patent medicine, setting a second threshold M, sorting the multi-scale feature information entropy, and obtaining the top M sorted features; The multi-scale information entropy of each component of the Chinese patent medicine is calculated by the following process: ； in, represents the information entropy of m scales, Indicates j The weight of each scale, m represents the total number of scales; S5. The multi-scale information entropy of each component of the Chinese patent medicine calculated in S4 is used as a feature to perform machine deep learning, and the up-regulation, down-regulation or no effect of the Chinese patent medicine on each gene target is used as the target variable; Construct drug-target-function network.

2. The analysis method according to claim 1, characterized in that The data information defined by the scale includes the scale of molecular structure, the scale of drug efficacy target, the scale of drug target gene interaction, and the scale of gene function and metabolic pathway.

3. The analysis method according to claim 1, characterized in that The data preprocessing in step S2 includes cleaning, formatting and normalizing the data.

4. The analysis method according to claim 3, characterized in that The multi-scale information entropy of each component of the Chinese patent medicine includes structural entropy, target entropy, interaction entropy and functional entropy.

5. The analysis method according to claim 1, characterized in that The Chinese patent medicine includes compound danshen tablets.

6. A method for constructing a model of network pharmacology of traditional Chinese medicine based on multi-scale information entropy, characterized in that: The model building method comprises the following steps: S1. Data collection, including information on components of Chinese patent medicines, small molecule compounds, target genes and interaction information, gene functions and metabolic pathways; S2, preprocessing the collected data; S3. Calculate the feature information entropy of each scale for each data, and select the top N ranked features of each scale. The feature information entropy calculation process of each scale is as follows: ； in, Indicates j Information entropy of scale, Indicates j The scale i The probability of an event occurring, n represents the total number of events; S4. Calculate the information entropy of each scale of each component of the Chinese patent medicine, sort the multi-scale feature information entropy, and obtain the features of the top M sorts. The calculation process is: ； in, represents the information entropy of m scales, Indicates j The weight of each scale, m represents the total number of scales; S5. The multi-scale information entropy of each component of the Chinese patent medicine calculated in S4 is used as a feature to build a deep learning model, and the up-regulation, down-regulation or no effect of each target is used as the target variable; the Chinese patent medicine and the target are randomly combined to obtain the sample data of each combination, and the up-regulated and down-regulated targets are selected as positive samples in the Chinese patent medicine component target database, and other unknown labels are used as negative samples.

7. The model building method according to claim 6, characterized in that: The input data of the model are information entropies of various scales calculated based on the components of Chinese patent medicines, including structural entropy, target entropy, interaction entropy, and functional entropy; as well as data on the differential gene expression levels of each sample after intervention with Chinese patent medicines; the output data of the model are the effects of Chinese patent medicines on the target.

Images