A disease target gene discovery framework is provided to detect susceptible genes and causal variants, which consists of two modules, including colocalization analysis and tissue propensity analysis.
Colocalization analysis can be performed with xQTLbiolinks by providing their own GWAS summary statistics data. All we need to prepare include three parts:
library(data.table)
library(xQTLbiolinks)
library(stringr)Download and load an example file of summary statistics dataset (GRCh38). We perform colocalization analysis in Brain - Cerebellum.
gwasDF <- fread("http://raw.githubusercontent.com/dingruofan/exampleData/master/gwas/AD/gwasDFsub.txt")
tissueSiteDetail="Brain - Cerebellum"In this example, a data.table object of 16538 (rows) x 5 (cols) is loaded. Five columns are required (arbitrary column names is supported, but columns must be as the following order):
Col 1. “variants” (character), , using an rsID (e.g. “rs11966562”);
Col 2. “chromosome” (character), one of the chromosome from chr1-chr22;
Col 3. “position” (integer), genome position of snp;
Col 4. “P-value” (numeric);
Col 5. “MAF” (numeric). Allel frequency;
Col 6. “beta” (numeric). effect size.
Col 7. “se” (numeric). standard error.
head(gwasDF)#> rsid chrom position pValue AF beta se
#> 1: rs13120565 4 10702513 5.66e-10 0.6429 0.01825 0.00294
#> 2: rs4697781 4 10703243 8.94e-10 0.6446 0.01803 0.00294
#> 3: rs4697779 4 10701187 5.74e-09 0.3231 -0.01747 0.00300
#> 4: rs4697780 4 10701381 5.95e-09 0.3231 -0.01746 0.00300
#> 5: rs4547789 4 10702891 1.46e-08 0.3197 -0.01703 0.00300
#> 6: rs11726285 4 10700944 1.47e-08 0.3214 -0.01702 0.00301Sentinel SNPs can be detected using xQTLanalyze_getSentinelSnp with the arguments p-value < 5e-8 and SNP-to-SNP distance > 10e6 bp. We recommend converting the genome version of the GWAS file to GRCh38 if that is GRCh37 (run with parameter: genomeVersion="grch37"; grch37To38=TRUE, and package rtracklayeris required).
sentinelSnpDF <- xQTLanalyze_getSentinelSnp(gwasDF, pValueThreshold = 5e-08)After filtering, a sentinel SNP with P-value<5e-8 is detected in this example:
sentinelSnpDF#> rsid chr position pValue maf beta se
#> 1: rs13120565 chr4 10702513 5.66e-10 0.6429 0.01825 0.00294Trait genes are genes that located in the range of 1Mb (default, can be changed with parameter detectRange) of sentinel SNPs. Every gene within 1Mb of sentinel SNPs is searched by fuction xQTLanalyze_getTraits. Besides, In order to reduce the number of trait genes and thus reduce the running time, we take the overlap of eGenes and trait genes as the final output of the function xQTLanalyze_getTraits.
traitsAll <- xQTLanalyze_getTraits(sentinelSnpDF, detectRange=1e6, tissueSiteDetail=tissueSiteDetail)Totally, 3 associations between 3 traits genes and 1 SNPs are detected
traitsAll#> chromosome geneStart geneEnd geneStrand geneSymbol gencodeId
#> 1: chr4 11393150 11429765 - HS3ST1 ENSG00000002587.9
#> 2: chr4 10486395 10684865 - CLNK ENSG00000109684.14
#> 3: chr4 10068089 10073019 - RP11-448G15.3 ENSG00000261490.1
#> rsid position pValue maf
#> 1: rs13120565 10702513 5.66e-10 0.6429
#> 2: rs13120565 10702513 5.66e-10 0.6429
#> 3: rs13120565 10702513 5.66e-10 0.6429For a single trait gene, like CLNK in above table, colocalization analysis (using coloc method) can be performed with:
output <- xQTLanalyze_coloc(gwasDF, "CLNK", tissueSiteDetail=tissueSiteDetail) # using gene symboloutput is a list, including three parts: coloc_Out_summary, gwasEqtlInfo, and gwasEqtlInfo.
output$coloc_Out_summary#> nsnps PP.H0.abf PP.H1.abf PP.H2.abf PP.H3.abf PP.H4.abf traitGene
#> 1: 7107 7.097893e-11 1.32221e-07 3.890211e-06 0.00625302 0.993743 CLNK
#> candidate_snp SNP.PP.H4
#> 1: rs13120565 0.5328849For multiple trait genes, a for loop or lapply function can be used to get all genes’ outputs (using both coloc and hyprcoloc methods).
outputs <- rbindlist(lapply( unique(traitsAll$gencodeId), function(x){ # using gencode ID.
xQTLanalyze_coloc(gwasDF, x, tissueSiteDetail=tissueSiteDetail, method = "Both")$colocOut }))outputs is a data.table that combined all results of coloc_Out_summary of all genes.
outputs#> traitGene nsnps PP.H0.abf PP.H1.abf PP.H2.abf PP.H3.abf
#> 1: ENSG00000002587.9 6452 1.730175e-05 3.218430e-02 6.603361e-05 0.12198838
#> 2: ENSG00000109684.14 7107 7.097893e-11 1.322210e-07 3.890211e-06 0.00625302
#> 3: ENSG00000261490.1 6601 5.287051e-05 9.848309e-02 4.801374e-04 0.89435622
#> PP.H4.abf candidate_snp SNP.PP.H4 hypr_posterior hypr_regional_prob
#> 1: 0.84574398 rs13120565 0.4140146 0.5685 0.9694
#> 2: 0.99374296 rs13120565 0.5328849 0.9793 0.9999
#> 3: 0.00662768 rs13120565 0.4219650 0.0000 0.0101
#> hypr_candidate_snp hypr_posterior_explainedBySnp
#> 1: rs13120565 0.2726
#> 2: rs13120565 0.4747
#> 3: rs13120565 0.4118For the potential casual gene ENSG00000109684.14 (PP4=0.9937 & hypr_posterior=0.9793) with candidate SNP rs13120565 (SNP.PP.H4=0.5328849 & hypr_posterior_explainedBySnp=0.4747), We merge the variants of GWAS and eQTL by rsid.
eqtlAsso <- xQTLdownload_eqtlAllAsso(gene="ENSG00000109684.14",
tissueLabel = tissueSiteDetail)
gwasEqtldata <- merge(gwasDF, eqtlAsso[,.(rsid=snpId, position=pos, maf, pValue)],
by=c("rsid", "position"), suffixes = c(".gwas",".eqtl"))Visualization of p-value distribution and comparison of the signals of GWAS and eQTL:
xQTLvisual_locusCompare(gwasEqtldata[,.(rsid, pValue.eqtl)],
gwasEqtldata[,.(rsid, pValue.gwas)], legend_position = "bottomright")Locuszoom plot of GWAS signals:
xQTLvisual_locusZoom(gwasEqtldata[,.(rsid, chrom, position, pValue.gwas)], legend=FALSE)Locuszoom plot of eQTL signals:
xQTLvisual_locusZoom(gwasEqtldata[,.(rsid, chrom, position, pValue.eqtl)],
highlightSnp = "rs13120565", legend=FALSE)Violin plot of normalized exprssion of eQTL (rs13120565-ENSG00000187323.11):
xQTLvisual_eqtlExp("rs13120565", "ENSG00000109684.14", tissueSiteDetail = tissueSiteDetail)We can also combine locuscompare and locuszoom plot using xQTLvisual_locusCombine:
xQTLvisual_locusCombine(gwasEqtldata[,c("rsid","chrom", "position", "pValue.gwas", "pValue.eqtl")],
highlightSnp="rs13120565")Colocalized loci should show a general pattern where SNPs in high LD will show strong associations with expression levels of the colocalized gene, and the eQTL associations will weaken for SNPs in lower LD. This pattern of the eQTL varies among different tissues / cell-types, the strength of which indicates the power of the regulatory effect of the variant. We can easily distinguish this patten using Tissue Propensity Analysis, and uncover the tissue / cell-type that the variant can play the most significant regulatory role in.
propensityRes <- xQTLanalyze_propensity( gene="ENSG00000109684.14", variantName="rs13120565",
tissueLabels = c("Brain - Cerebellar Hemisphere",
"Brain - Cerebellum", "Treg memory", "Colon - Transverse",
"CD4+ T cell", "Kidney - Cortex", "Uterus", "Esophagus - Mucosa", "LCL") )A p-value<0.05 that calculated via permutation test indicates there exists a significant correlation between the correlation coefficient of LD and eQTL significance.
propensityRes$tissuePropensity#> tissue_label study_accession tissue pValue_eQTL
#> 1: Brain - Cerebellar Hemisphere GTEx_V8 UBER_0002037 2.34596e-06
#> 2: Treg memory Schmiedel_2018 CL_0002678 2.25235e-05
#> 3: Kidney - Cortex GTEx_V8 UBER_0001225 1.36117e-01
#> 4: LCL GENCORD EFO_0005292 5.85962e-01
#> 5: Esophagus - Mucosa GTEx_V8 UBER_0006920 7.95494e-01
#> 6: Uterus GTEx_V8 UBER_0000995 9.78559e-01
#> 7: CD4+ T cell BLUEPRINT CL_0000624 8.51989e-01
#> 8: Colon - Transverse GTEx_V8 UBER_0001157 7.02369e-01
#> 9: Brain - Cerebellum GTEx_V8 UBER_0002037 1.37779e-13
#> pValue_propensity
#> 1: 0.000999001
#> 2: 0.027972028
#> 3: 0.606393606
#> 4: 1.000000000
#> 5: 0.950049950
#> 6: 0.716283716
#> 7: 0.381618382
#> 8: 0.244755245
#> 9: 0.000999001using a heatmap plot to visualize the result:
xQTLvisual_qtlPropensity(propensityRes)For applying xQTLbiolinks to a whole case study, please find this Document.