Drug target screening method based on Mendel randomization principle analysis
Through the drug target screening method based on Mendel's randomization principle, the problem of lack of credible methods in the existing technology to explore the therapeutic effects and side effects of proteins as drug targets is solved, and the effective confirmation of the causal relationship between proteins and diseases is achieved and Verification of the potential role of protein inhibitors.
Patent Information
- Application Number
- CN202411466000.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-21
- Publication Date
- 2025-02-11
AI Technical Summary
The prior art lacks a simple and reliable method to explore the therapeutic effects, potential side effects and therapeutic effects of proteins as drug targets on diseases.
The drug target screening method based on Mendelian randomization principle was used to obtain pQTL data and GWAS data, and the causal relationship was analyzed using the two-sample Mendelian randomization method, and the results were verified through positive control experiments and multiple preferred analytical methods.
This method can effectively confirm the causal relationship between proteins and diseases, determine the potential therapeutic effects and side effects of protein inhibitors on the disease, and provides a new research direction and basis for drug development.
Smart Images
Figure CN119418758A_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the technical field of biomedicine, and in particular relates to a method for screening drug targets. Background Art
[0002] Common arthritis includes rheumatoid arthritis, osteoarthritis, psoriatic arthritis, juvenile idiopathic arthritis, ankylosing spondylitis, and gouty arthritis, which impose a heavy burden on society and individuals, and lack safe and effective drugs.
[0003] S100A4 is expressed in many cells related to arthritis, such as macrophages and fibroblasts. Intracellular S100A4 expression will bind to many targets, causing the infiltration of inflammatory cells and the release of inflammatory factors. Extracellular S100A4 can activate the NF-κB pathway, thereby causing inflammation.
[0004] Studies have shown that S100A4 is associated with many types of arthritis, but the causal relationship between S100A4 and various types of arthritis is unknown, and S100A4 has not been explored as a drug target for the treatment of arthritis.
[0005] Thanks to the rapid development of pQTL data, we can use pQTL data to simulate the level of a certain plasma protein. In addition, we can use Mendelian randomization to explore the causal relationship between a certain plasma protein and a group of diseases. And we can use positive control experiments to test whether the exposure simulation is a suppressor or an enhancer.
[0006] At present, the preliminary verification of drug targets can be completed through molecular biology experiments and animal experiments, while the verification of the effect of a protein inhibitor on the disease can be completed through randomized controlled experiments. The former is simpler, but not credible enough, while the latter is very credible, but there are problems such as taking a long time and consuming manpower and material resources.
[0007] Currently, there is no simple, easy and reliable method to explore the therapeutic effect, potential side effects and therapeutic effects of a protein as a drug target. The present invention provides a simple and effective method that can be used to evaluate the therapeutic effect, potential other therapeutic effects and side effects of a protein inhibitor, and uses the effect of S100A4 inhibitor on arthritis as an example. Summary of the invention
[0008] The present invention aims to provide a method for screening drug targets based on Mendelian randomization principle analysis.
[0009] The drug target screening method provided by the present invention analyzes the relationship between a certain plasma protein and a group of diseases according to Mendel's randomization theory, determines it as a drug target for a certain disease, and verifies it. The specific steps are:
[0010] Step 1: Obtain the pQTL data of the protein [1-3] and the GWAS data of a group of diseases [4-9];
[0011] Step 2, using the two-sample Mendelian randomization method
[10] to analyze the pQTL data and GWAS data to obtain preliminary results of the causal protein;
[0012] Step 3, positive control analysis was used to confirm that the effect simulated using pQTL data was a suppressor effect;
[0013] Step 4: Analyze the preliminary results using replication analysis methods
[11] , Steiger filtering analysis methods
[12] , sensitivity analysis methods [13-15], pQTL and eQTL overlap assessment methods
[16] , and Bayesian co-localization analysis methods
[17] to confirm causal relationships;
[0014] Step 5: Use the phenotypic-Mendelian randomization method
[18] to analyze the potential therapeutic effects and side effects of the protein inhibitor.
[0015] Further:
[0016] Step 2 includes the following sub-steps:
[0017] Step 2.1, let the SNP with strong correlation with plasma protein in pQTL data as instrumental variable, the selected plasma protein as exposure factor, and the GWAS data of a group of selected diseases as outcome factor; based on the two-sample Mendelian randomization theory, evaluate the causal relationship between the protein and the disease;
[0018] Step 2.2, using the Benjamini–Hochberg method
[19] to modify the P threshold of the above results, screen which diseases the selected plasma proteins are related to, and obtain preliminary results of the relationship between the proteins and diseases;
[0019] Step 3 includes the following sub-steps:
[0020] Step 3.1, select GWAS data of diseases that can be treated by the protein inhibitor;
[0021] Step 3.2: Use the SNP determined in step 2.1 as the instrumental variable, the selected protein as the exposure factor, and the selected disease as the outcome factor; use the two-sample Mendelian randomization design theory to evaluate the causal relationship between the selected protein and the selected disease;
[0022] Preferably, in step 3.2, the result is greater than 0 and the p-value is less than 0.05, indicating that the simulated value is a suppressor.
[0023] Preferably, step 4 includes the following sub-steps:
[0024] Step 4.1, using replication analysis to verify the causal relationship between protein and disease;
[0025] Step 4.2, using Steiger filtering analysis to verify whether the causal relationship direction between the single SNP in the preliminary screening results and the selected disease is correct;
[0026] Step 4.3, sensitivity analysis methods were used, including MR Egger, weighted median, weighted mode, and simple mode methods to estimate the consistent causal effects obtained by different methods, Cochrane's Q test to assess whether there is heterogeneity between SNPs, MR-Egger and MR-PRESSO methods to test whether there is pleiotropy, and the leave-one-out method
[20] to observe whether the direction of the results changes after removing each SNP separately;
[0027] Step 4.4, using the method of evaluating the overlap between pQTL and eQTL to strengthen the credibility of the causal relationship between the protein and the disease;
[0028] In step 4.5, Bayesian co-localization analysis was used to calculate the posterior probability that the protein and the disease were driven by the same SNP based on their association with a certain genomic region, further verifying the causal relationship between the protein and the disease.
[0029] Step 5 includes the following sub-steps:
[0030] Step 5.1: Let the protein inhibitor verified in step 3 be the exposure factor, the SNP corresponding to the protein inhibitor be the instrumental variable, and the common diseases and their GWAS data be the outcome factors; perform full phenotype Mendelian randomization analysis with the help of a two-sample Mendelian randomization strategy to clarify the causal relationship between the protein inhibitor and common diseases;
[0031] Step 5.2, using the Bonferroni method to modify the P threshold of the above results, screen and exclude proteins and gene loci that have significant causal relationships with common diseases other than the arthritis disease;
[0032] Preferably, the common disease GWAS data are sourced from two databases, UK biobank and FinnGen, to verify the reliability of the results.
[0033] The process of the method of the present invention can be written in code to form a software system, which constitutes a screening system for proteins analyzed based on Mendel's randomization theory as drug targets, thereby realizing fully automated screening.
[0034] That is, the present invention also provides a screening system for proteins as drug targets based on Mendelian randomization analysis, which specifically includes five modules: a module for acquiring pQTL data of the protein and GWAS data of a group of diseases; a module for double-sample Mendelian randomization analysis of pQTL data and GWAS data; a module for positive control analysis to determine inhibitor effects; a module for analyzing preliminary results to confirm causal relationships; and a module for analyzing the potential therapeutic effects and side effects of the protein inhibitor. The five modules perform the operations of the five steps in the screening method to achieve the screening of drug targets.
[0035] The technical features and effects of the present invention mainly include:
[0036] By using Mendelian randomization and three plasma cis-pQTL data, we confirmed the causal relationship between S100A4 and rheumatoid arthritis and osteoarthritis, providing a basis for S100A4 as a biomarker for these two types of arthritis.
[0037] By using positive control studies, we determined that the instrumental variables extracted from the pQTL data were simulated S100A4 inhibitors, thus confirming that S100A4 inhibitors have therapeutic effects on rheumatoid arthritis and osteoarthritis, providing new research directions for drug targets for these two types of arthritis;
[0038] A variety of optimal analysis methods were used, including replication analysis, Steiger filtering analysis, sensitivity analysis, pQTL and eQTL overlap analysis, and co-localization analysis, to determine the reliability of the causal relationship obtained;
[0039] By using the phenotypic-Mendelian randomization method, the potential therapeutic effects and side effects of S100A4 inhibitors were explored, including potential harm to patients with uterine fibroids, but potential benefits to patients with gastroesophageal reflux disease, varicose veins of the lower extremities, and other non-inflammatory diseases of the vulva and perineum, providing a reference for drug precautions when S100A4 inhibitors are used clinically;
[0040] The present invention provides a method for exploring the causal relationship between a protein and a group of diseases and the protein inhibitor as a drug target. According to this method, the causal relationship between any protein and a group of diseases, the therapeutic effect of the protein inhibitor on the group of diseases as a drug target, and the potential therapeutic effect and side effects of the protein inhibitor are explored. BRIEF DESCRIPTION OF THE DRAWINGS
[0041] Figure 1 This is a schematic diagram of the process of exploring the causal relationship between S100A4 inhibitors and common arthritis in Example 1. DETAILED DESCRIPTION
[0042] The present invention is further described below by means of specific embodiments.
[0043] Example 1, Screening of S100A4 as a common arthritis drug target.
[0044] This example analyzes the relationship between S100A4 and arthritis. Figure 1 As shown, the specific steps are as follows:
[0045] Step 1: Obtain relevant pQTL data and GWAS data.
[0046] 1. Three pQTL data of plasma S100A4 (exposure data):
[0047] The sources of pQTL data were: a pQTL study conducted by Sun et al. in a cohort of 54,219 UK Biobank participants; a study by Ferkingstad et al. in a cohort of 35,559 Icelanders who participated in the Icelandic Cancer Project or various projects of deCODE genetics; and a study by Gudjonsson et al. in a cohort of 5368 persons of Northern European ancestry who participated in the AGES-Reykjavík study.
[0048] After the pQTL data were analyzed, the metal software was used to perform meta-analysis on these three pQTLs to obtain the final S100A4 pQTL data.
[0049] 2. GWAS data (outcome data) of 6 common arthritis:
[0050] Most of the outcome GWAS data used in the preliminary study came from the GWAS Catalog website (https: / / www.ebi.ac.uk / gwas / ), and the access numbers for rheumatoid arthritis, psoriatic arthritis, juvenile idiopathic arthritis, and gout were GCST90132223, GCST90243956, GCST90010715, and GCST008971, respectively.
[0051] The ankylosing spondylitis GWAS data were obtained from the Finnish database with the access number M13_ANKYLOSPON, and the osteoarthritis GWAS data were obtained from the website provided in the literature (https: / / msk.hugeamp.org / downloads.html).
[0052] The outcome GWAS data used in the replication analysis were from the GWAS Catalog website, UK Biobank, and the Finnish database. The access numbers for rheumatoid arthritis, psoriatic arthritis, juvenile idiopathic arthritis, ankylosing spondylitis, and gouty osteoarthritis were ieu-a-832, M13_PSORIARTH, JUVEN_ARTHR, ukb-a-88, GCST90038687, and GCST007093, respectively.
[0053] 3. GWAS data of amyotrophic lateral sclerosis (positive control group):
[0054] The ALS GWAS data used were from the GWAS Catalog website, with access numbers GCST90132223 and GCST005647, and sample sizes of 138,086 and 80,610, respectively.
[0055] All the above pQTL data and GWAS data are from public databases and can be accessed and downloaded.
[0056] Step 2: In order to explore the causal relationship between plasma S100A4 protein and six common arthritis, Mendelian randomization analysis was performed.
[0057] 1. Selection of instrumental variables:
[0058] In the pQTL data after the meta-analysis of S100A4, SNPs located within ±100 kb of the S100A4 locus, with a P value less than 5e-8 and an F value (calculated as the square of the effect value divided by the square of the standard error) greater than 10 were screened. The clump function in the PLINK package in R language was used to ensure linkage equilibrium. These SNPs were selected as instrumental variables for subsequent Mendelian randomization analysis, see Table 1.
[0059] Table 1
[0060]
[0061] 2. Perform Mendelian randomization analysis
[0062] The above selected SNPs were used as instrumental variables to represent the exposure factor of plasma S100A4 protein, and the corresponding arthritis GWAS data were used as the outcome factor. The causal relationship between plasma S100A4 protein and six common arthritis was evaluated using the two-sample Mendelian randomization theory (Table 2). The main method adopted was the inverse variance weighting method.
[0063] 3. Correct the results
[0064] The Benjamini–Hochberg method was used to correct the P threshold of the above results. If the corrected P value was less than 0.05, it indicated that plasma S100A4 protein was related to this type of arthritis.
[0065] The above process can be completed under R software compilation, and the packages used are TwoSampleMR and plinkbinr packages.
[0066] Table 2
[0067]
[0068] Step 3: To explore whether the simulated effect is an inhibitor, we will perform a positive control analysis
[0069] The extracted SNPs are selected as instrumental variables to simulate exposure factors, and the ALS GWAS data are selected as outcome factors for Mendelian randomization. If the result shows that the effect value is less than 0, it means that the selected SNP simulates the S100A4 inhibitor effect. This further indicates that the S100A4 inhibitor has a potential therapeutic effect on the above-mentioned arthritis related to it (Table 2). The above process can be completed in the R language environment, and the package used is the TwoSampleMR package.
[0070] Step 4: To make the results more reliable, we performed replication analysis, Steiger filtering analysis, sensitivity analysis, pQTL and eQTL overlap, and co-localization analysis.
[0071] 1. Copy analysis
[0072] Mendelian randomization is performed again using outcome GWAS data and positive control GWAS data from different data sources. If the results are consistent with the preliminary results, the results are credible; otherwise, the results are unreliable.
[0073] 2. Steiger filtering analysis
[0074] Each SNP was examined to see whether its correlation with plasma S100A4 protein level was significantly greater than its correlation with the outcome. If so, this would indicate that the Mendelian randomization study results were an exposure-outcome relationship, rather than an outcome-exposure relationship.
[0075] 3. Sensitivity analysis
[0076] MR Egger, weighted median, weighted mode, and simple mode were used to detect whether consistent results could be obtained under different methods. Cochrane's Q test was used to evaluate whether there was heterogeneity among SNPs. MR-Egger and MR-PRESSO methods were used to test whether there was pleiotropy. The leave-one-out method was used to see whether the direction of the results changed after removing each SNP.
[0077] 4. Overlap between pQTL and eQTL
[0078] The method of evaluating the degree of overlap between pQTL and eQTL is used to see whether there is any overlap between the SNPs extracted from the pQTL data and the SNPs extracted from the eQTLGen and GTEx databases. If there is overlap and the direction of the effect values is consistent, the relationship between protein and disease is more credible.
[0079] 5. Bayesian co-localization analysis
[0080] The posterior probability that the protein and the mental illness are driven by the same SNP based on their association with a certain genomic region is calculated. If the posterior probability is greater than 0.80, it means that the protein and the disease can be driven by the same SNP, and their causal relationship is more credible. If the posterior probability is less than 0.80, it is impossible to determine their causal relationship, which is even less credible.
[0081] The above process can be completed in R software, using the TwoSampleMR and coloc packages.
[0082] Step 5: To explore the potential therapeutic effects and side effects of S100A4 inhibitors when S100A4 is used as a drug target, we used the phenotypic-Mendelian randomization method.
[0083] 1. Extract SNPs and perform Mendelian randomization
[0084] We extracted phenotypes other than arthritis and amyotrophic lateral sclerosis from the UK Biobank, including 514 phenotypes, and then used Mendelian randomization analysis. If one common SNP was finally obtained, the Waldratio method was used. If two or more SNPs were finally obtained, the inverse variance weighting method was used (Table 3).
[0085] 2. Correct the results
[0086] The Bonferroni method was used to correct the P threshold of the above results. If the P value of the corrected result was less than 0.05, it was proved that the S100A4 inhibitor was related to it (Table 3).
[0087] 3. Replication Analysis
[0088] The same phenotype as the positive phenotype was extracted from the Finnish database, and the above steps were repeated, and the positive result obtained was the final result (Table 3).
[0089] The above process can be implemented in R language software, and the package used is the TwoSampleMR package.
[0090] Table 3
[0091]
[0092] References
[0093] 1.Gudjonsson A, Gudmundsdottir V, Axelsson GT, et al. A genome-wide association study of serum proteins reveals shared loci with common diseases. Nat Commun. 2022;13(1):480.
[0095] 2. Sun BB, Chiou J, Traylor M, et al. Plasma proteomic associations with genetics and health in the UK Biobank. Nature. 2023; 622(7982): 329-338.
[0096] 3. Ferkingstad E, Sulem P, Atlason BA, et al. Large-scale integration of the plasma proteome with genetics and disease. Nat Genet. 2021; 53(12): 1712-1721.
[0097] 4.Ishigaki K, Sakaue S, Terao C, et al. Multi-ancestry genome-wide association analyzes identify novel genetic mechanisms in rheumatoidarthritis. Nat Genet.
[0098] 2022;54(11):1640-1651.
[0099] 5.Boer CG,Hatzikotoulas K,Southam L,et al.Deciphering osteoarthritisgenetics across 826,690 individuals from 9 populations.Cell.2021;184(18):4784-4818.e17.6.Soomro M,Stadler M,Dand N,et al.Comparative Genetic Analysisof Psoriatic Arthritis and Psoriasis for the Discovery of Genetic RiskFactors and Risk Prediction Modeling.Arthritis Rheumatol.2022;74(9):1535-1543.
[0100] 7.López-Isac E,Smith SL,Marion MC,et al.Combined genetic analysis ofjuvenile idiopathic arthritis clinical subtypes identifies novel risk loci,target genes and key regulatory mechanisms.Ann Rheum Dis.2021;80(3):321-328.
[0101] 8.Kurki MI,Karjalainen J,Palta P,et al.FinnGen provides geneticinsights from a well-phenotyped isolated population.Nature.2023;613(7944):508-518.
[0102] 9.Tin A,Marten J,Halperin Kuhns VL,et al.Target genes,variants,tissues and transcriptional pathways influencing human serum urate levels.NatGenet.2019;51(10):1459-1474.
[0104] 10.Sheehan NA,Didelez V,Burton PR,Tobin MD.Mendelian randomisationand causal inference in observational epidemiology.PLoS Med.2008;5(8):e177.
[0105] 11.Xu Q,Ni JJ,Han BX,et al.Causal Relationship Between Gut Microbiotaand Autoimmune Diseases:A Two-Sample Mendelian Randomization Study.FrontImmunol.2022;12:746998.
[0107] 12.Hemani G,Tilling K,Davey Smith G.Orienting the causal relationshipbetween imprecisely measured traits using GWAS summary data.PLoS Genet.
[0108] 2017;13(11):e1007081.
[0109] 13.Bowden J,Davey Smith G,Haycock PC,Burgess S.Consistent Estimationin Mendelian Randomization with Some Invalid Instruments Using a WeightedMedian Estimator.
[0110] Genet Epidemiol.2016;40(4):304-314.
[0111] 14.Schmidt AF,Dudbridge F.Mendelian randomization with Eggerpleiotropy correction and weakly informative Bayesian priors.Int JEpidemiol.2018;47(4):1217-1228.15.Verbanck M,Chen CY,Neale B,Do R.Detectionof widespread horizontal pleiotropy in causal relationships inferred fromMendelian randomization between complex traits and diseases.Nat Genet.2018;50(5):693-698.
[0112] 16.Zhang Y,Xie J,Wen S,et al.Evaluating the causal effect ofcirculating proteome on the risk of osteoarthritis-related traits.Ann RheumDis.
[0113] 2023;82(12):1606-1617.
[0114] 17.Deng YT,Ou YN,Wu BS,et al.Identifying causal genes for depressionvia integration of the proteome and transcriptome from brain and blood.MolPsychiatry.2022;27(6):2849-2857.
[0116] 18.Constantinescu AE,Bull CJ,Goudswaard LJ,et al.A phenome-wideapproach to identify causal risk factors for deep vein thrombosis.BMC MedGenomics. 2023;16(1):284.
[0118] 19.Shu X,Zhou Q,Sun X,et al.Associations between circulating proteinsand risk of breast cancer by intrinsic subtypes:a Mendelian randomisationanalysis.Br J Cancer.2022;127(8):1507-1514.
[0119] 20.Luo S,Li W,Li Q,et al.Causal effects of gut microbiota on the riskof periodontitis:a two-sample Mendelian randomization study.Front Cell InfectMicrobiol.2023;13:1160993。
Claims
1. A drug target screening method based on Mendelian randomization analysis, characterized in that: According to Mendel's randomization theory, the relationship between a certain plasma protein and a group of diseases is analyzed, and it is determined as a drug target for a certain disease and verified. The specific steps are: Step 1, obtain the pQTL data of the protein and the GWAS data of a group of diseases; Step 2, using a two-sample Mendelian randomization method to analyze the pQTL data and GWAS data to obtain preliminary results of the causal protein; Step 3, positive control analysis was used to confirm that the effect simulated using pQTL data was a suppressor effect; Step 4, using replication analysis method, Steiger filtering analysis method, sensitivity analysis method, pQTL and eQTL overlap evaluation method, and Bayesian co-localization analysis method to analyze the preliminary results to confirm the causal relationship; Step 5, using phenotypic-Mendelian randomization method to analyze the potential therapeutic effects and side effects of the protein inhibitor.
2. The screening method according to claim 1, characterized in that Step 2 includes the following sub-steps: Step 2.1, let the SNPs with strong correlation with plasma proteins in the pQTL data be instrumental variables, the selected plasma proteins be exposure factors, and the GWAS data of a group of selected diseases be outcome factors; Based on the two-sample Mendelian randomization theory, the causal relationship between the protein and the disease was evaluated; Step 2.2, use the Benjamini–Hochberg method to correct the P threshold of the above results, screen which diseases the selected plasma proteins are related to, and obtain preliminary results of the relationship between the proteins and diseases.
3. The screening method according to claim 2, characterized in that Step 3 includes the following sub-steps: Step 3.1, select GWAS data of diseases that can be treated by the protein inhibitor; Step 3.2, using the SNPs determined in step 2.1 as instrumental variables, the selected proteins as exposure factors, and the selected diseases as outcome factors; Two-sample Mendelian randomization design theory was used to evaluate the causal relationship between the selected protein and the selected disease; If the result is greater than 0 and the p-value is less than 0.05, it means that the simulated protein is a suppressor.
4. The screening method according to claim 3, characterized in that Step 4 includes the following sub-steps: Step 4.1, using replication analysis to verify the causal relationship between protein and disease; Step 4.2, using the Steiger filtering analysis method to verify whether the causal relationship direction between the single SNP in the preliminary screening results and the selected disease is correct; Step 4.3, sensitivity analysis methods were used, including the use of MR Egger, weighted median, weighted mode, and simple mode methods to estimate the consistent causal effects obtained by different methods, the use of Cochrane's Q test to assess whether there is heterogeneity between SNPs, the use of MR-Egger and MR-PRESSO methods to test whether there is pleiotropy, and the use of the leave-one-out method to observe whether the direction of the results changes after removing each SNP separately; Step 4.4, using the method of evaluating the overlap between pQTL and eQTL to strengthen the credibility of the causal relationship between the protein and the disease; In step 4.5, Bayesian co-localization analysis was used to calculate the posterior probability that the protein and the disease were driven by the same SNP based on their association with a certain genomic region, further verifying the causal relationship between the protein and the disease.
5. The screening method according to claim 4, characterized in that Step 5 includes the following sub-steps: Step 5.1: Let the protein inhibitor verified in step 3 be the exposure factor, the SNP corresponding to the protein inhibitor be the instrumental variable, and the common diseases and their GWAS data be the outcome factors; perform full phenotype Mendelian randomization analysis with the help of a two-sample Mendelian randomization strategy to clarify the causal relationship between the protein inhibitor and common diseases; Step 5.2, using the Bonferroni method to modify the P threshold of the above results, screen and exclude proteins and their gene loci that have a significant causal relationship with common diseases other than the arthritis disease.
6. A drug target screening system based on the screening method according to any one of claims 1 to 5, characterized in that: It includes five modules: a module for acquiring protein pQTL data and a group of disease GWAS data; a module for two-sample Mendelian randomization analysis of pQTL data and GWAS data; Positive control analysis identified suppressor effector modules; The preliminary results are analyzed to confirm the causal relationship module; the potential therapeutic effect and side effect analysis module of the protein inhibitor; the five modules perform the five steps in the screening method to achieve the screening of drug targets.