Simulation Results using three Tissues (uncorrelated)
shengqian
2023-10-13
Last updated: 2023-10-13
Checks: 6 1
Knit directory: cTWAS_analysis/
This reproducible R Markdown analysis was created with workflowr (version 1.7.0). The Checks tab describes the reproducibility checks that were applied when the results were created. The Past versions tab lists the development history.
The R Markdown is untracked by Git. To know which version of the R Markdown file created these results, you’ll want to first commit it to the Git repo. If you’re still working on the analysis, you can ignore this warning. When you’re finished, you can run wflow_publish to commit the R Markdown file and build the HTML.
Great job! The global environment was empty. Objects defined in the global environment can affect the analysis in your R Markdown file in unknown ways. For reproduciblity it’s best to always run the code in an empty environment.
The command set.seed(20211220) was run prior to running the code in the R Markdown file. Setting a seed ensures that any results that rely on randomness, e.g. subsampling or permutations, are reproducible.
Great job! Recording the operating system, R version, and package versions is critical for reproducibility.
Nice! There were no cached chunks for this analysis, so you can be confident that you successfully produced the results during this run.
Great job! Using relative paths to the files within your workflowr project makes it easier to run your code on other machines.
Great! You are using Git for version control. Tracking code development and connecting the code version to the results is critical for reproducibility.
The results in this page were generated with repository version 68eacf2. See the Past versions tab to see a history of the changes made to the R Markdown and HTML files.
Note that you need to be careful to ensure that all relevant files for the analysis have been committed to Git prior to generating the results (you can use wflow_publish or wflow_git_commit). workflowr only checks the R Markdown file, but you know if there are other scripts or data files that it depends on. Below is the status of the Git repository when the results were generated:
Ignored files:
Ignored: .Rhistory
Ignored: .ipynb_checkpoints/
Untracked files:
Untracked: LDL_LDLR_locus1.pdf
Untracked: LDL_TEME199_genetrack.pdf
Untracked: LDL_TEME199_locus.pdf
Untracked: Proposal plots.R
Untracked: RGS14.pdf
Untracked: RNF186.pdf
Untracked: Rplots.pdf
Untracked: SCZ_annotation.xlsx
Untracked: SLC8B1.pdf
Untracked: analysis/.ipynb_checkpoints/
Untracked: analysis/OxHB.Rmd
Untracked: analysis/simulation_3tissues_correlated.Rmd
Untracked: analysis/simulation_3tissues_uncorrelated.Rmd
Untracked: cache/
Untracked: code/.ipynb_checkpoints/
Untracked: code/OxHb_out/OxHb_Kidney_Cortex.err
Untracked: code/OxHb_out/OxHb_Kidney_Cortex.out
Untracked: code/OxHb_out/OxHb_Liver.err
Untracked: code/OxHb_out/OxHb_Liver.out
Untracked: code/OxHb_out/OxHb_Pancreas.err
Untracked: code/OxHb_out/OxHb_Pancreas.out
Untracked: code/OxHb_out/OxHb_Whole_Blood.err
Untracked: code/OxHb_out/OxHb_Whole_Blood.out
Untracked: data/.ipynb_checkpoints/
Untracked: data/FUMA_output/
Untracked: data/GO_Terms/
Untracked: data/GTEx_Analysis_v8_eQTL.tar
Untracked: data/G_list.RData
Untracked: data/IBD_ME/
Untracked: data/LB/
Untracked: data/LDL/
Untracked: data/LDL_E_S/
Untracked: data/LDL_E_S_M/
Untracked: data/LDL_M/
Untracked: data/LDL_S/
Untracked: data/OxHb/
Untracked: data/PGC3_SCZ_wave3_public.v2.tsv
Untracked: data/Predictive_Models/
Untracked: data/Supplementary Table 15 - MAGMA.xlsx
Untracked: data/Supplementary Table 20 - Prioritised Genes.xlsx
Untracked: data/UKBB/
Untracked: data/UKBB_SNPs_Info.text
Untracked: data/WhiteBlood_E/
Untracked: data/WhiteBlood_E_M/
Untracked: data/WhiteBlood_E_S_M/
Untracked: data/WhiteBlood_E_S_M_PC/
Untracked: data/WhiteBlood_M/
Untracked: data/WhiteBlood_M_compare/
Untracked: data/WhiteBlood_M_enet/
Untracked: data/cpg_annot.RData
Untracked: data/eqtl/
Untracked: data/gencode.v26.GRCh38.genes.gtf
Untracked: data/gene_OMIM.txt
Untracked: data/gene_pip_0.8.txt
Untracked: data/gwas_sumstats/
Untracked: data/magma.genes.out
Untracked: data/mashr_Heart_Atrial_Appendage.db
Untracked: data/mashr_sqtl/
Untracked: data/mqtl/
Untracked: data/notes.txt
Untracked: data/scz_2018.RDS
Untracked: data/summary_known_genes_annotations.xlsx
Untracked: data/test/
Untracked: hist.pdf
Untracked: ld_pip.pdf
Untracked: submit.sh
Untracked: temp_LDR/
Untracked: test-B1.snpgwas.txt
Unstaged changes:
Deleted: analysis/Atrial_Fibrillation_Heart_Atrial_Appendage.Rmd
Deleted: analysis/Atrial_Fibrillation_Heart_Left_Ventricle.Rmd
Deleted: analysis/Autism_Brain_Amygdala.Rmd
Deleted: analysis/Autism_Brain_Anterior_cingulate_cortex_BA24.Rmd
Deleted: analysis/Autism_Brain_Caudate_basal_ganglia.Rmd
Deleted: analysis/Autism_Brain_Cerebellar_Hemisphere.Rmd
Deleted: analysis/Autism_Brain_Cerebellum.Rmd
Deleted: analysis/Autism_Brain_Cortex.Rmd
Deleted: analysis/Autism_Brain_Frontal_Cortex_BA9.Rmd
Deleted: analysis/Autism_Brain_Hippocampus.Rmd
Deleted: analysis/Autism_Brain_Hypothalamus.Rmd
Deleted: analysis/Autism_Brain_Nucleus_accumbens_basal_ganglia.Rmd
Deleted: analysis/Autism_Brain_Putamen_basal_ganglia.Rmd
Deleted: analysis/Autism_Brain_Spinal_cord_cervical_c-1.Rmd
Deleted: analysis/Autism_Brain_Substantia_nigra.Rmd
Deleted: analysis/BMI_Brain_Amygdala.Rmd
Deleted: analysis/BMI_Brain_Amygdala_S.Rmd
Deleted: analysis/BMI_Brain_Anterior_cingulate_cortex_BA24.Rmd
Deleted: analysis/BMI_Brain_Anterior_cingulate_cortex_BA24_S.Rmd
Deleted: analysis/BMI_Brain_Caudate_basal_ganglia.Rmd
Deleted: analysis/BMI_Brain_Caudate_basal_ganglia_S.Rmd
Deleted: analysis/BMI_Brain_Cerebellar_Hemisphere.Rmd
Deleted: analysis/BMI_Brain_Cerebellar_Hemisphere_S.Rmd
Deleted: analysis/BMI_Brain_Cerebellum.Rmd
Deleted: analysis/BMI_Brain_Cerebellum_S.Rmd
Deleted: analysis/BMI_Brain_Cortex.Rmd
Deleted: analysis/BMI_Brain_Cortex_S.Rmd
Deleted: analysis/BMI_Brain_Frontal_Cortex_BA9.Rmd
Deleted: analysis/BMI_Brain_Frontal_Cortex_BA9_S.Rmd
Deleted: analysis/BMI_Brain_Hippocampus.Rmd
Deleted: analysis/BMI_Brain_Hippocampus_S.Rmd
Deleted: analysis/BMI_Brain_Hypothalamus.Rmd
Deleted: analysis/BMI_Brain_Hypothalamus_S.Rmd
Deleted: analysis/BMI_Brain_Nucleus_accumbens_basal_ganglia.Rmd
Deleted: analysis/BMI_Brain_Nucleus_accumbens_basal_ganglia_S.Rmd
Deleted: analysis/BMI_Brain_Putamen_basal_ganglia.Rmd
Deleted: analysis/BMI_Brain_Putamen_basal_ganglia_S.Rmd
Deleted: analysis/BMI_Brain_Spinal_cord_cervical_c-1.Rmd
Deleted: analysis/BMI_Brain_Spinal_cord_cervical_c-1_S.Rmd
Deleted: analysis/BMI_Brain_Substantia_nigra.Rmd
Deleted: analysis/BMI_Brain_Substantia_nigra_S.Rmd
Deleted: analysis/BMI_S_results.Rmd
Deleted: analysis/Glucose_Adipose_Subcutaneous.Rmd
Deleted: analysis/Glucose_Adipose_Visceral_Omentum.Rmd
Modified: analysis/index.Rmd
Deleted: code/OxHb_out/OxHb_Artery_Aorta.err
Deleted: code/OxHb_out/OxHb_Artery_Aorta.out
Modified: code/OxHb_out/OxHb_Artery_Coronary.err
Modified: code/OxHb_out/OxHb_Artery_Coronary.out
Modified: code/OxHb_out/OxHb_Artery_Tibial.out
Modified: code/OxHb_out/OxHb_Heart_Atrial_Appendage.err
Modified: code/OxHb_out/OxHb_Heart_Atrial_Appendage.out
Modified: code/OxHb_out/OxHb_Heart_Left_Ventricle.out
Modified: code/OxHb_out/OxHb_Lung.err
Modified: code/OxHb_out/OxHb_Lung.out
Deleted: code/WhiteBlood_M_ener_out/WhiteBlood_WholeBlood.out
Deleted: code/White_Blood_M_out/White_Blood_BreastMammary.err
Deleted: code/White_Blood_M_out/White_Blood_BreastMammary.out
Deleted: code/White_Blood_M_out/White_Blood_ColonTransverse.err
Deleted: code/White_Blood_M_out/White_Blood_ColonTransverse.out
Deleted: code/White_Blood_M_out/White_Blood_KidneyCortex.err
Deleted: code/White_Blood_M_out/White_Blood_KidneyCortex.out
Deleted: code/White_Blood_M_out/White_Blood_Lung.err
Deleted: code/White_Blood_M_out/White_Blood_Lung.out
Deleted: code/White_Blood_M_out/White_Blood_MuscleSkeletal.err
Deleted: code/White_Blood_M_out/White_Blood_MuscleSkeletal.out
Deleted: code/White_Blood_M_out/White_Blood_Ovary.err
Deleted: code/White_Blood_M_out/White_Blood_Ovary.out
Deleted: code/White_Blood_M_out/White_Blood_Prostate.err
Deleted: code/White_Blood_M_out/White_Blood_Prostate.out
Deleted: code/White_Blood_M_out/White_Blood_Testis.err
Deleted: code/White_Blood_M_out/White_Blood_Testis.out
Deleted: code/White_Blood_M_out/White_Blood_WholeBlood.err
Deleted: code/White_Blood_M_out/White_Blood_WholeBlood.out
Deleted: code/run_IBD_ctwas_rss_LDR_ME.R
Modified: code/run_OxHb_analysis.sh
Modified: code/run_OxHb_ctwas_rss_LDR.R
Note that any generated files, e.g. HTML, png, CSS, etc., are not included in this status report because it is ok for generated content to have uncommitted changes.
There are no past versions. Publish this analysis with wflow_publish() to start tracking its development.
Simulation 1: Two causal tissues with equal high PVE
Separate effect size parameters
For the cTWAS analysis, each tissue had its own prior inclusion parameter end effect size parameter.
results_dir <- "/project2/xinhe/shengqian/cTWAS/cTWAS_simulation/simulation_uncorrelated/"
runtag = "ukb-s80.45-3_uncorr"
configtag <- 2
simutags <- paste(1, 1:3, sep = "-")
thin <- 0.1
sample_size <- 45000
PIP_threshold <- 0.8results_df <- data.frame(simutag=as.character(),
n_causal=as.integer(),
n_causal_combined=as.integer(),
n_detected_pip=as.integer(),
n_detected_pip_in_causal=as.integer(),
n_detected_comb_pip=as.integer(),
n_detected_comb_pip_in_causal=as.integer(),
pve_snp=as.numeric(),
pve_weight1=as.numeric(),
pve_weight2=as.numeric(),
pve_weight3=as.numeric(),
prior_weight1=as.numeric(),
prior_weight2=as.numeric(),
prior_weight3=as.numeric(),
prior_var_snp=as.numeric(),
prior_var_weight1=as.numeric(),
prior_var_weight2=as.numeric(),
prior_var_weight3=as.numeric())
for (i in 1:length(simutags)){
simutag <- simutags[i]
#load genes with true simulated effect
load(paste0(results_dir, runtag, "_simu", simutag, "-pheno.Rd"))
true_genes <- unlist(sapply(1:22, function(x){phenores$batch[[x]]$id.cgene}))
true_genes_combined <- unique(sapply(true_genes, function(x){unlist(strsplit(x, "[|]"))[1]}))
#load cTWAS results
ctwas_res <- data.table::fread(paste0(results_dir, runtag, "_simu", simutag, "_config", configtag, "_LDR.susieIrss.txt"))
ctwas_gene_res <- ctwas_res[ctwas_res$type!="SNP",]
#number of causal genes
n_causal <- length(true_genes)
n_causal_combined <- length(true_genes_combined)
#number of gene+tissue combinations with cTWAS PIP > threshold
n_ctwas_genes <- sum(ctwas_gene_res$susie_pip > PIP_threshold)
#number of cTWAS genes that are causal
n_causal_detected <- sum(ctwas_gene_res$id[ctwas_gene_res$susie_pip > PIP_threshold] %in% true_genes)
#collapse gene+tissues to genes and compute combined PIP
ctwas_gene_res$gene <- sapply(ctwas_gene_res$id, function(x){unlist(strsplit(x,"[|]"))[1]})
ctwas_gene_res_combined <- aggregate(ctwas_gene_res$susie_pip, by=list(ctwas_gene_res$gene), FUN=sum)
colnames(ctwas_gene_res_combined) <- c("gene", "pip_combined")
#number of genes with combined PIP > threshold
n_ctwas_genes_combined <- sum(ctwas_gene_res_combined$pip_combined > PIP_threshold)
#number of cTWAS genes using combined PIP that are causal
n_causal_detected_combined <- sum(ctwas_gene_res_combined$gene[ctwas_gene_res_combined$pip_combined > PIP_threshold] %in% true_genes_combined)
#collect number of SNPs analyzed by cTWAS
ctwas_res_s1 <- data.table::fread(paste0(results_dir, runtag, "_simu", simutag, "_config", configtag, "_LDR.s1.susieIrss.txt"))
n_snps <- sum(ctwas_res_s1$type=="SNP")/thin
rm(ctwas_res_s1)
#load estimated parameters
load(paste0(results_dir, runtag, "_simu", simutag, "_config", configtag, "_LDR.s2.susieIrssres.Rd"))
#estimated group prior (all iterations)
estimated_group_prior_all <- group_prior_rec
estimated_group_prior_all["SNP",] <- estimated_group_prior_all["SNP",]*thin #adjust parameter to account for thin argument
#estimated group prior variance (all iterations)
estimated_group_prior_var_all <- group_prior_var_rec
#set group size
group_size <- c(table(ctwas_gene_res$type), structure(n_snps, names="SNP"))
group_size <- group_size[rownames(estimated_group_prior_all)]
#estimated group PVE (all iterations)
estimated_group_pve_all <- estimated_group_prior_var_all*estimated_group_prior_all*group_size/sample_size #check PVE calculation
results_current <- data.frame(simutag=as.character(simutag),
n_causal=as.integer(n_causal),
n_causal_combined=as.integer(n_causal_combined),
n_detected_pip=as.integer(n_ctwas_genes),
n_detected_pip_in_causal=as.integer(n_causal_detected),
n_detected_comb_pip=as.integer(n_ctwas_genes_combined),
n_detected_comb_pip_in_causal=as.integer(n_causal_detected_combined),
pve_snp=as.numeric(rev(estimated_group_pve_all["SNP",])[1]),
pve_weight1=as.numeric(rev(estimated_group_pve_all["Liver_harmonized",])[1]),
pve_weight2=as.numeric(rev(estimated_group_pve_all["Lung_harmonized",])[1]),
pve_weight3=as.numeric(rev(estimated_group_pve_all["Brain_Hippocampus_harmonized",])[1]),
prior_snp=as.numeric(rev(estimated_group_prior_var_all["SNP",])[1]),
prior_weight1=as.numeric(rev(estimated_group_prior_all["Liver_harmonized",])[1]),
prior_weight2=as.numeric(rev(estimated_group_prior_all["Lung_harmonized",])[1]),
prior_weight3=as.numeric(rev(estimated_group_prior_all["Brain_Hippocampus_harmonized",])[1]),
prior_var_snp=as.numeric(rev(estimated_group_prior_var_all["SNP",])[1]),
prior_var_weight1=as.numeric(rev(estimated_group_prior_var_all["Liver_harmonized",])[1]),
prior_var_weight2=as.numeric(rev(estimated_group_prior_var_all["Lung_harmonized",])[1]),
prior_var_weight3=as.numeric(rev(estimated_group_prior_var_all["Brain_Hippocampus_harmonized",])[1]))
results_df <- rbind(results_df, results_current)
}
#results using PIP threshold (gene+tissue)
results_df[,c("simutag", "n_causal", "n_detected_pip", "n_detected_pip_in_causal")] simutag n_causal n_detected_pip n_detected_pip_in_causal
1 1-1 211 48 18
2 1-2 228 81 25
3 1-3 199 52 17#mean percent causal using PIP > 0.8
sum(results_df$n_detected_pip_in_causal)/sum(results_df$n_detected_pip)[1] 0.3314917#results using combined PIP threshold
results_df[,c("simutag", "n_causal_combined", "n_detected_comb_pip", "n_detected_comb_pip_in_causal")] simutag n_causal_combined n_detected_comb_pip n_detected_comb_pip_in_causal
1 1-1 211 52 28
2 1-2 228 91 36
3 1-3 199 58 21#mean percent causal using combined PIP > 0.8
sum(results_df$n_detected_comb_pip_in_causal)/sum(results_df$n_detected_comb_pip)[1] 0.4228856#prior inclusion and mean prior inclusion
results_df[,c(which(colnames(results_df)=="simutag"), setdiff(grep("prior", names(results_df)), grep("prior_var", names(results_df))))] simutag prior_snp prior_weight1 prior_weight2 prior_weight3
1 1-1 8.905386 0.011687231 0.007045722 0.004102420
2 1-2 9.136868 0.009150253 0.004080808 0.008522311
3 1-3 9.079034 0.006371810 0.007416565 0.016031927colMeans(results_df[,setdiff(grep("prior", names(results_df)), grep("prior_var", names(results_df)))]) prior_snp prior_weight1 prior_weight2 prior_weight3
9.040429271 0.009069765 0.006181032 0.009552219 #prior variance and mean prior variance
results_df[,c(which(colnames(results_df)=="simutag"), grep("prior_var", names(results_df)))] simutag prior_var_snp prior_var_weight1 prior_var_weight2 prior_var_weight3
1 1-1 8.905386 4.740432 10.075945 35.451613
2 1-2 9.136868 16.152947 25.129552 23.308393
3 1-3 9.079034 22.318028 9.712662 5.500348colMeans(results_df[,grep("prior_var", names(results_df))]) prior_var_snp prior_var_weight1 prior_var_weight2 prior_var_weight3
9.040429 14.403803 14.972720 21.420118 #PVE and mean PVE
results_df[,c(which(colnames(results_df)=="simutag"), grep("pve", names(results_df)))] simutag pve_snp pve_weight1 pve_weight2 pve_weight3
1 1-1 0.2678360 0.009095864 0.01536589 0.02569394
2 1-2 0.2628193 0.024266060 0.02219614 0.03509331
3 1-3 0.2785189 0.023347104 0.01559149 0.01557868colMeans(results_df[,grep("pve", names(results_df))]) pve_snp pve_weight1 pve_weight2 pve_weight3
0.26972472 0.01890301 0.01771784 0.02545531 #store results for figure
plot_df <- rbind(plot_df,
data.frame(simutag=results_df$simutag,
method="cTWAS G - Ind Var",
count=results_df$n_detected_comb_pip-results_df$n_detected_comb_pip_in_causal,
ifcausal=F))
plot_df <- rbind(plot_df,
data.frame(simutag=results_df$simutag,
method="cTWAS G - Ind Var",
count=results_df$n_detected_comb_pip_in_causal,
ifcausal=T))Individual tissue analyses
For the cTWAS analysis, each tissue was analyzed individually and the results were combined.
results_dir <- "/project2/xinhe/shengqian/cTWAS/cTWAS_simulation/simulation_uncorrelated/"
runtag = "ukb-s80.45-3_uncorr"
configtag <- 2
simutags <- paste(1, 1:3, sep = "-")
thin <- 0.1
sample_size <- 45000
PIP_threshold <- 0.8
results_df <- data.frame(simutag=as.character(),
n_causal_combined=as.integer(),
n_detected_weight1=as.integer(),
n_detected_in_causal_weight1=as.integer(),
n_detected_weight2=as.integer(),
n_detected_in_causal_weight2=as.integer(),
n_detected_weight3=as.integer(),
n_detected_in_causal_weight3=as.integer(),
n_detected_combined=as.integer(),
n_detected_in_causal_combined=as.integer())
for (i in 1:length(simutags)){
simutag <- simutags[i]
#load genes with true simulated effect
load(paste0(results_dir, runtag, "_simu", simutag, "-pheno.Rd"))
true_genes <- unlist(sapply(1:22, function(x){phenores$batch[[x]]$id.cgene}))
true_genes_combined <- unique(sapply(true_genes, function(x){unlist(strsplit(x, "[|]"))[1]}))
#number of causal genes
n_causal_combined <- length(true_genes_combined)
#load cTWAS results for weight1
ctwas_res <- data.table::fread(paste0(results_dir, runtag, "_simu", simutag, "_config", configtag, "_LDR_weight1.susieIrss.txt"))
ctwas_gene_res <- ctwas_res[ctwas_res$type!="SNP",]
#number of genes with cTWAS PIP > threshold in weight1
ctwas_genes_weight1 <- ctwas_gene_res$id[ctwas_gene_res$susie_pip > PIP_threshold]
n_ctwas_genes_weight1 <- length(ctwas_genes_weight1)
n_causal_detected_weight1 <- sum(ctwas_genes_weight1 %in% true_genes_combined)
#load cTWAS results for weight2
ctwas_res <- data.table::fread(paste0(results_dir, runtag, "_simu", simutag, "_config", configtag, "_LDR_weight2.susieIrss.txt"))
ctwas_gene_res <- ctwas_res[ctwas_res$type!="SNP",]
#number of genes with cTWAS PIP > threshold in weight1
ctwas_genes_weight2 <- ctwas_gene_res$id[ctwas_gene_res$susie_pip > PIP_threshold]
n_ctwas_genes_weight2 <- length(ctwas_genes_weight2)
n_causal_detected_weight2 <- sum(ctwas_genes_weight2 %in% true_genes_combined)
#load cTWAS results for weight3
ctwas_res <- data.table::fread(paste0(results_dir, runtag, "_simu", simutag, "_config", configtag, "_LDR_weight3.susieIrss.txt"))
ctwas_gene_res <- ctwas_res[ctwas_res$type!="SNP",]
#number of genes with cTWAS PIP > threshold in weight3
ctwas_genes_weight3 <- ctwas_gene_res$id[ctwas_gene_res$susie_pip > PIP_threshold]
n_ctwas_genes_weight3 <- length(ctwas_genes_weight3)
n_causal_detected_weight3 <- sum(ctwas_genes_weight3 %in% true_genes_combined)
#combined analysis
ctwas_genes_combined <- unique(c(ctwas_genes_weight1, ctwas_genes_weight2, ctwas_genes_weight3))
n_ctwas_genes_combined <- length(ctwas_genes_combined)
n_causal_detected_combined <- sum(ctwas_genes_combined %in% true_genes_combined)
results_current <- data.frame(simutag=as.character(simutag),
n_causal_combined=as.integer(n_causal_combined),
n_detected_weight1=as.integer(n_ctwas_genes_weight1),
n_detected_in_causal_weight1=as.integer(n_causal_detected_weight1),
n_detected_weight2=as.integer(n_ctwas_genes_weight2),
n_detected_in_causal_weight2=as.integer(n_causal_detected_weight2),
n_detected_weight3=as.integer(n_ctwas_genes_weight3),
n_detected_in_causal_weight3=as.integer(n_causal_detected_weight3),
n_detected_combined=as.integer(n_ctwas_genes_combined),
n_detected_in_causal_combined=as.integer(n_causal_detected_combined))
results_df <- rbind(results_df, results_current)
}
#results using weight1
results_df[,c("simutag", colnames(results_df)[grep("weight1", colnames(results_df))])] simutag n_detected_weight1 n_detected_in_causal_weight1
1 1-1 33 13
2 1-2 48 23
3 1-3 48 20#mean percent causal using PIP > 0.8 for weight1
sum(results_df$n_detected_in_causal_weight1)/sum(results_df$n_detected_weight1)[1] 0.4341085#results using weight2
results_df[,c("simutag", colnames(results_df)[grep("weight2", colnames(results_df))])] simutag n_detected_weight2 n_detected_in_causal_weight2
1 1-1 20 7
2 1-2 49 18
3 1-3 59 22#mean percent causal using PIP > 0.8 for weight1
sum(results_df$n_detected_in_causal_weight2)/sum(results_df$n_detected_weight2)[1] 0.3671875#results using weight3
results_df[,c("simutag", colnames(results_df)[grep("weight3", colnames(results_df))])] simutag n_detected_weight3 n_detected_in_causal_weight3
1 1-1 33 19
2 1-2 49 20
3 1-3 52 21#mean percent causal using PIP > 0.8 for weight3
sum(results_df$n_detected_in_causal_weight3)/sum(results_df$n_detected_weight3)[1] 0.4477612#results using combined analysis
results_df[,c("simutag", colnames(results_df)[grep("combined", colnames(results_df))])] simutag n_causal_combined n_detected_combined n_detected_in_causal_combined
1 1-1 211 71 33
2 1-2 228 100 41
3 1-3 199 121 44#mean percent causal using PIP > 0.8 for weight3
sum(results_df$n_detected_in_causal_combined)/sum(results_df$n_detected_combined)[1] 0.4041096#store results for figure
plot_df <- rbind(plot_df,
data.frame(simutag=results_df$simutag,
method="cTWAS Weight1",
count=results_df$n_detected_weight1-results_df$n_detected_in_causal_weight1,
ifcausal=F))
plot_df <- rbind(plot_df,
data.frame(simutag=results_df$simutag,
method="cTWAS Weight1",
count=results_df$n_detected_in_causal_weight1,
ifcausal=T))
plot_df <- rbind(plot_df,
data.frame(simutag=results_df$simutag,
method="cTWAS Weight2",
count=results_df$n_detected_weight2-results_df$n_detected_in_causal_weight2,
ifcausal=F))
plot_df <- rbind(plot_df,
data.frame(simutag=results_df$simutag,
method="cTWAS Weight2",
count=results_df$n_detected_in_causal_weight2,
ifcausal=T))
plot_df <- rbind(plot_df,
data.frame(simutag=results_df$simutag,
method="cTWAS Weight3",
count=results_df$n_detected_weight3-results_df$n_detected_in_causal_weight3,
ifcausal=F))
plot_df <- rbind(plot_df,
data.frame(simutag=results_df$simutag,
method="cTWAS Weight3",
count=results_df$n_detected_in_causal_weight3,
ifcausal=T))
plot_df <- rbind(plot_df,
data.frame(simutag=results_df$simutag,
method="cTWAS Union",
count=results_df$n_detected_combined-results_df$n_detected_in_causal_combined,
ifcausal=F))
plot_df <- rbind(plot_df,
data.frame(simutag=results_df$simutag,
method="cTWAS Union",
count=results_df$n_detected_in_causal_combined,
ifcausal=T))Figure
For the cTWAS analysis, each tissue was analyzed individually and the results were combined.
plot_df$method <- factor(plot_df$method, levels=c("TWAS", "cTWAS G+T", "cTWAS G", "cTWAS G - Ind Var", "cTWAS Union", "cTWAS Weight1", "cTWAS Weight2", "cTWAS Weight3"))
library(ggpubr)Loading required package: ggplot2colset = c("#ebebeb", "#fb8072")
ggbarplot(plot_df,
x = "method",
y = "count",
add = "mean_se",
fill = "ifcausal",
legend = "none",
ylab="Count",
xlab="",
palette = colset) + grids(linetype = "dashed") + theme(axis.text.x = element_text(angle = 45, vjust = 1, hjust=1))
plot_df_2 <- plot_df
plot_df_2$method %in% c("TWAS", "cTWAS Union", "cTWAS Weight1", "cTWAS Weight2", "cTWAS Weight3") [1] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE
[13] TRUE TRUE TRUE TRUE TRUE TRUE FALSE FALSE FALSE FALSE FALSE FALSE
[25] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
[37] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUEplot_df_2$ifcausal <- as.character(plot_df_2$ifcausal+1)
plot_df_2$count[plot_df_2$method=="cTWAS G" & plot_df_2$ifcausal=="2"] <- plot_df_2$count[plot_df_2$method=="cTWAS G" & plot_df_2$ifcausal=="2"] - plot_df_2$count[plot_df_2$method=="cTWAS G+T" & plot_df_2$ifcausal=="2"]
plot_df_2 <- rbind(plot_df_2, data.frame(simutag=simutags,
method="cTWAS G",
count=plot_df_2$count[plot_df_2$method=="cTWAS G+T" & plot_df_2$ifcausal=="2"],
ifcausal="3"))
plot_df_3 <- plot_df_2
plot_df_2 <- plot_df_2[plot_df_2$method %in% c("TWAS", "cTWAS Union", "cTWAS Weight1", "cTWAS Weight2", "cTWAS Weight3"),]
plot_df_2$method <- droplevels(plot_df_2$method)
levels(plot_df_2$method) <- c("TWAS", "cTWAS Union", "cTWAS Primary", "cTWAS Secondary", "cTWAS Null")
plot_df_2$ifcausal[plot_df_2$ifcausal=="1"] <- "FP"
plot_df_2$ifcausal[plot_df_2$ifcausal=="2"] <- "TP"
colset = c("#ebebeb", "#fb8072", "#8dd3c7")
ggbarplot(plot_df_2,
x = "method",
y = "count",
add = "mean_se",
fill = "ifcausal",
legend = "right",
ylab="Count",
xlab="",
palette = colset) + grids(linetype = "dashed") + theme(axis.text.x = element_text(angle = 45, vjust = 1, hjust=1)) + guides(fill=guide_legend(title=""))
plot_df_3 <- plot_df_3[plot_df_3$method %in% c("cTWAS G", "cTWAS Union"),]
plot_df_3$method <- droplevels(plot_df_3$method)
levels(plot_df_3$method) <- c("cTWAS Joint", "cTWAS Union")
plot_df_3$method <- relevel(plot_df_3$method, "cTWAS Union")
plot_df_3$ifcausal[plot_df_3$ifcausal=="1"] <- "FP"
plot_df_3$ifcausal[plot_df_3$ifcausal=="2"] <- "TP"
plot_df_3$ifcausal[plot_df_3$ifcausal=="3"] <- "TP and Tissue Identified\nby Joint Analysis"
ggbarplot(plot_df_3,
x = "method",
y = "count",
add = "mean_se",
fill = "ifcausal",
legend = "right",
ylab="Count",
xlab="",
palette = colset) + grids(linetype = "dashed") + theme(axis.text.x = element_text(angle = 45, vjust = 1, hjust=1)) + guides(fill=guide_legend(title=""))
Simulation 2: Two causal tissues with equal low PVE
Separate effect size parameters
For the cTWAS analysis, each tissue had its own prior inclusion parameter and effect size parameter.
results_dir <- "/project2/xinhe/shengqian/cTWAS/cTWAS_simulation/simulation_uncorrelated/"
runtag = "ukb-s80.45-3_uncorr"
configtag <- 2
simutags <- paste(2, 1:3, sep = "-")
thin <- 0.1
sample_size <- 45000
PIP_threshold <- 0.8results_df <- data.frame(simutag=as.character(),
n_causal=as.integer(),
n_causal_combined=as.integer(),
n_detected_pip=as.integer(),
n_detected_pip_in_causal=as.integer(),
n_detected_comb_pip=as.integer(),
n_detected_comb_pip_in_causal=as.integer(),
pve_snp=as.numeric(),
pve_weight1=as.numeric(),
pve_weight2=as.numeric(),
pve_weight3=as.numeric(),
prior_weight1=as.numeric(),
prior_weight2=as.numeric(),
prior_weight3=as.numeric(),
prior_var_snp=as.numeric(),
prior_var_weight1=as.numeric(),
prior_var_weight2=as.numeric(),
prior_var_weight3=as.numeric())
for (i in 1:length(simutags)){
simutag <- simutags[i]
#load genes with true simulated effect
load(paste0(results_dir, runtag, "_simu", simutag, "-pheno.Rd"))
true_genes <- unlist(sapply(1:22, function(x){phenores$batch[[x]]$id.cgene}))
true_genes_combined <- unique(sapply(true_genes, function(x){unlist(strsplit(x, "[|]"))[1]}))
#load cTWAS results
ctwas_res <- data.table::fread(paste0(results_dir, runtag, "_simu", simutag, "_config", configtag, "_LDR.susieIrss.txt"))
ctwas_gene_res <- ctwas_res[ctwas_res$type!="SNP",]
#number of causal genes
n_causal <- length(true_genes)
n_causal_combined <- length(true_genes_combined)
#number of gene+tissue combinations with cTWAS PIP > threshold
n_ctwas_genes <- sum(ctwas_gene_res$susie_pip > PIP_threshold)
#number of cTWAS genes that are causal
n_causal_detected <- sum(ctwas_gene_res$id[ctwas_gene_res$susie_pip > PIP_threshold] %in% true_genes)
#collapse gene+tissues to genes and compute combined PIP
ctwas_gene_res$gene <- sapply(ctwas_gene_res$id, function(x){unlist(strsplit(x,"[|]"))[1]})
ctwas_gene_res_combined <- aggregate(ctwas_gene_res$susie_pip, by=list(ctwas_gene_res$gene), FUN=sum)
colnames(ctwas_gene_res_combined) <- c("gene", "pip_combined")
#number of genes with combined PIP > threshold
n_ctwas_genes_combined <- sum(ctwas_gene_res_combined$pip_combined > PIP_threshold)
#number of cTWAS genes using combined PIP that are causal
n_causal_detected_combined <- sum(ctwas_gene_res_combined$gene[ctwas_gene_res_combined$pip_combined > PIP_threshold] %in% true_genes_combined)
#collect number of SNPs analyzed by cTWAS
ctwas_res_s1 <- data.table::fread(paste0(results_dir, runtag, "_simu", simutag, "_config", configtag, "_LDR.s1.susieIrss.txt"))
n_snps <- sum(ctwas_res_s1$type=="SNP")/thin
rm(ctwas_res_s1)
#load estimated parameters
load(paste0(results_dir, runtag, "_simu", simutag, "_config", configtag, "_LDR.s2.susieIrssres.Rd"))
#estimated group prior (all iterations)
estimated_group_prior_all <- group_prior_rec
estimated_group_prior_all["SNP",] <- estimated_group_prior_all["SNP",]*thin #adjust parameter to account for thin argument
#estimated group prior variance (all iterations)
estimated_group_prior_var_all <- group_prior_var_rec
#set group size
group_size <- c(table(ctwas_gene_res$type), structure(n_snps, names="SNP"))
group_size <- group_size[rownames(estimated_group_prior_all)]
#estimated group PVE (all iterations)
estimated_group_pve_all <- estimated_group_prior_var_all*estimated_group_prior_all*group_size/sample_size #check PVE calculation
results_current <- data.frame(simutag=as.character(simutag),
n_causal=as.integer(n_causal),
n_causal_combined=as.integer(n_causal_combined),
n_detected_pip=as.integer(n_ctwas_genes),
n_detected_pip_in_causal=as.integer(n_causal_detected),
n_detected_comb_pip=as.integer(n_ctwas_genes_combined),
n_detected_comb_pip_in_causal=as.integer(n_causal_detected_combined),
pve_snp=as.numeric(rev(estimated_group_pve_all["SNP",])[1]),
pve_weight1=as.numeric(rev(estimated_group_pve_all["Liver_harmonized",])[1]),
pve_weight2=as.numeric(rev(estimated_group_pve_all["Lung_harmonized",])[1]),
pve_weight3=as.numeric(rev(estimated_group_pve_all["Brain_Hippocampus_harmonized",])[1]),
prior_snp=as.numeric(rev(estimated_group_prior_var_all["SNP",])[1]),
prior_weight1=as.numeric(rev(estimated_group_prior_all["Liver_harmonized",])[1]),
prior_weight2=as.numeric(rev(estimated_group_prior_all["Lung_harmonized",])[1]),
prior_weight3=as.numeric(rev(estimated_group_prior_all["Brain_Hippocampus_harmonized",])[1]),
prior_var_snp=as.numeric(rev(estimated_group_prior_var_all["SNP",])[1]),
prior_var_weight1=as.numeric(rev(estimated_group_prior_var_all["Liver_harmonized",])[1]),
prior_var_weight2=as.numeric(rev(estimated_group_prior_var_all["Lung_harmonized",])[1]),
prior_var_weight3=as.numeric(rev(estimated_group_prior_var_all["Brain_Hippocampus_harmonized",])[1]))
results_df <- rbind(results_df, results_current)
}
#results using PIP threshold (gene+tissue)
results_df[,c("simutag", "n_causal", "n_detected_pip", "n_detected_pip_in_causal")] simutag n_causal n_detected_pip n_detected_pip_in_causal
1 2-1 66 53 11
2 2-2 80 61 14
3 2-3 60 62 5#mean percent causal using PIP > 0.8
sum(results_df$n_detected_pip_in_causal)/sum(results_df$n_detected_pip)[1] 0.1704545#results using combined PIP threshold
results_df[,c("simutag", "n_causal_combined", "n_detected_comb_pip", "n_detected_comb_pip_in_causal")] simutag n_causal_combined n_detected_comb_pip n_detected_comb_pip_in_causal
1 2-1 66 52 14
2 2-2 80 65 21
3 2-3 60 71 9#mean percent causal using combined PIP > 0.8
sum(results_df$n_detected_comb_pip_in_causal)/sum(results_df$n_detected_comb_pip)[1] 0.2340426#prior inclusion and mean prior inclusion
results_df[,c(which(colnames(results_df)=="simutag"), setdiff(grep("prior", names(results_df)), grep("prior_var", names(results_df))))] simutag prior_snp prior_weight1 prior_weight2 prior_weight3
1 2-1 8.849442 0.002382323 0.005985903 0.002692681
2 2-2 7.819633 0.001938417 0.003274174 0.006660638
3 2-3 7.749635 0.006701771 0.003545606 0.005927356colMeans(results_df[,setdiff(grep("prior", names(results_df)), grep("prior_var", names(results_df)))]) prior_snp prior_weight1 prior_weight2 prior_weight3
8.139570018 0.003674170 0.004268561 0.005093558 #prior variance and mean prior variance
results_df[,c(which(colnames(results_df)=="simutag"), grep("prior_var", names(results_df)))] simutag prior_var_snp prior_var_weight1 prior_var_weight2 prior_var_weight3
1 2-1 8.849442 57.84209 5.971779 19.66389
2 2-2 7.819633 50.68076 7.027209 21.18784
3 2-3 7.749635 10.02814 22.283548 21.99606colMeans(results_df[,grep("prior_var", names(results_df))]) prior_var_snp prior_var_weight1 prior_var_weight2 prior_var_weight3
8.13957 39.51700 11.76085 20.94927 #PVE and mean PVE
results_df[,c(which(colnames(results_df)=="simutag"), grep("pve", names(results_df)))] simutag pve_snp pve_weight1 pve_weight2 pve_weight3
1 2-1 0.2485026 0.02262346 0.007737129 0.009354252
2 2-2 0.2677475 0.01612890 0.004980021 0.024931998
3 2-3 0.2656854 0.01103378 0.017100991 0.023033534colMeans(results_df[,grep("pve", names(results_df))]) pve_snp pve_weight1 pve_weight2 pve_weight3
0.26064516 0.01659538 0.00993938 0.01910659 #store results for figure
plot_df <- rbind(plot_df,
data.frame(simutag=results_df$simutag,
method="cTWAS G - Ind Var",
count=results_df$n_detected_comb_pip-results_df$n_detected_comb_pip_in_causal,
ifcausal=F))
plot_df <- rbind(plot_df,
data.frame(simutag=results_df$simutag,
method="cTWAS G - Ind Var",
count=results_df$n_detected_comb_pip_in_causal,
ifcausal=T))Individual tissue analyses
For the cTWAS analysis, each tissue was analyzed individually and the results were combined.
results_dir <- "/project2/xinhe/shengqian/cTWAS/cTWAS_simulation/simulation_uncorrelated/"
runtag = "ukb-s80.45-3_uncorr"
configtag <- 2
simutags <- paste(2, 1:3, sep = "-")
thin <- 0.1
sample_size <- 45000
PIP_threshold <- 0.8
results_df <- data.frame(simutag=as.character(),
n_causal_combined=as.integer(),
n_detected_weight1=as.integer(),
n_detected_in_causal_weight1=as.integer(),
n_detected_weight2=as.integer(),
n_detected_in_causal_weight2=as.integer(),
n_detected_weight3=as.integer(),
n_detected_in_causal_weight3=as.integer(),
n_detected_combined=as.integer(),
n_detected_in_causal_combined=as.integer())
for (i in 1:length(simutags)){
simutag <- simutags[i]
#load genes with true simulated effect
load(paste0(results_dir, runtag, "_simu", simutag, "-pheno.Rd"))
true_genes <- unlist(sapply(1:22, function(x){phenores$batch[[x]]$id.cgene}))
true_genes_combined <- unique(sapply(true_genes, function(x){unlist(strsplit(x, "[|]"))[1]}))
#number of causal genes
n_causal_combined <- length(true_genes_combined)
#load cTWAS results for weight1
ctwas_res <- data.table::fread(paste0(results_dir, runtag, "_simu", simutag, "_config", configtag, "_LDR_weight1.susieIrss.txt"))
ctwas_gene_res <- ctwas_res[ctwas_res$type!="SNP",]
#number of genes with cTWAS PIP > threshold in weight1
ctwas_genes_weight1 <- ctwas_gene_res$id[ctwas_gene_res$susie_pip > PIP_threshold]
n_ctwas_genes_weight1 <- length(ctwas_genes_weight1)
n_causal_detected_weight1 <- sum(ctwas_genes_weight1 %in% true_genes_combined)
#load cTWAS results for weight2
ctwas_res <- data.table::fread(paste0(results_dir, runtag, "_simu", simutag, "_config", configtag, "_LDR_weight2.susieIrss.txt"))
ctwas_gene_res <- ctwas_res[ctwas_res$type!="SNP",]
#number of genes with cTWAS PIP > threshold in weight1
ctwas_genes_weight2 <- ctwas_gene_res$id[ctwas_gene_res$susie_pip > PIP_threshold]
n_ctwas_genes_weight2 <- length(ctwas_genes_weight2)
n_causal_detected_weight2 <- sum(ctwas_genes_weight2 %in% true_genes_combined)
#load cTWAS results for weight3
ctwas_res <- data.table::fread(paste0(results_dir, runtag, "_simu", simutag, "_config", configtag, "_LDR_weight3.susieIrss.txt"))
ctwas_gene_res <- ctwas_res[ctwas_res$type!="SNP",]
#number of genes with cTWAS PIP > threshold in weight3
ctwas_genes_weight3 <- ctwas_gene_res$id[ctwas_gene_res$susie_pip > PIP_threshold]
n_ctwas_genes_weight3 <- length(ctwas_genes_weight3)
n_causal_detected_weight3 <- sum(ctwas_genes_weight3 %in% true_genes_combined)
#combined analysis
ctwas_genes_combined <- unique(c(ctwas_genes_weight1, ctwas_genes_weight2, ctwas_genes_weight3))
n_ctwas_genes_combined <- length(ctwas_genes_combined)
n_causal_detected_combined <- sum(ctwas_genes_combined %in% true_genes_combined)
results_current <- data.frame(simutag=as.character(simutag),
n_causal_combined=as.integer(n_causal_combined),
n_detected_weight1=as.integer(n_ctwas_genes_weight1),
n_detected_in_causal_weight1=as.integer(n_causal_detected_weight1),
n_detected_weight2=as.integer(n_ctwas_genes_weight2),
n_detected_in_causal_weight2=as.integer(n_causal_detected_weight2),
n_detected_weight3=as.integer(n_ctwas_genes_weight3),
n_detected_in_causal_weight3=as.integer(n_causal_detected_weight3),
n_detected_combined=as.integer(n_ctwas_genes_combined),
n_detected_in_causal_combined=as.integer(n_causal_detected_combined))
results_df <- rbind(results_df, results_current)
}
#results using weight1
results_df[,c("simutag", colnames(results_df)[grep("weight1", colnames(results_df))])] simutag n_detected_weight1 n_detected_in_causal_weight1
1 2-1 30 9
2 2-2 39 11
3 2-3 24 4#mean percent causal using PIP > 0.8 for weight1
sum(results_df$n_detected_in_causal_weight1)/sum(results_df$n_detected_weight1)[1] 0.2580645#results using weight2
results_df[,c("simutag", colnames(results_df)[grep("weight2", colnames(results_df))])] simutag n_detected_weight2 n_detected_in_causal_weight2
1 2-1 25 4
2 2-2 22 4
3 2-3 39 5#mean percent causal using PIP > 0.8 for weight1
sum(results_df$n_detected_in_causal_weight2)/sum(results_df$n_detected_weight2)[1] 0.1511628#results using weight3
results_df[,c("simutag", colnames(results_df)[grep("weight3", colnames(results_df))])] simutag n_detected_weight3 n_detected_in_causal_weight3
1 2-1 27 7
2 2-2 34 13
3 2-3 32 3#mean percent causal using PIP > 0.8 for weight3
sum(results_df$n_detected_in_causal_weight3)/sum(results_df$n_detected_weight3)[1] 0.2473118#results using combined analysis
results_df[,c("simutag", colnames(results_df)[grep("combined", colnames(results_df))])] simutag n_causal_combined n_detected_combined n_detected_in_causal_combined
1 2-1 66 67 15
2 2-2 80 69 20
3 2-3 60 74 10#mean percent causal using PIP > 0.8 for weight3
sum(results_df$n_detected_in_causal_combined)/sum(results_df$n_detected_combined)[1] 0.2142857#store results for figure
plot_df <- rbind(plot_df,
data.frame(simutag=results_df$simutag,
method="cTWAS Weight1",
count=results_df$n_detected_weight1-results_df$n_detected_in_causal_weight1,
ifcausal=F))
plot_df <- rbind(plot_df,
data.frame(simutag=results_df$simutag,
method="cTWAS Weight1",
count=results_df$n_detected_in_causal_weight1,
ifcausal=T))
plot_df <- rbind(plot_df,
data.frame(simutag=results_df$simutag,
method="cTWAS Weight2",
count=results_df$n_detected_weight2-results_df$n_detected_in_causal_weight2,
ifcausal=F))
plot_df <- rbind(plot_df,
data.frame(simutag=results_df$simutag,
method="cTWAS Weight2",
count=results_df$n_detected_in_causal_weight2,
ifcausal=T))
plot_df <- rbind(plot_df,
data.frame(simutag=results_df$simutag,
method="cTWAS Weight3",
count=results_df$n_detected_weight3-results_df$n_detected_in_causal_weight3,
ifcausal=F))
plot_df <- rbind(plot_df,
data.frame(simutag=results_df$simutag,
method="cTWAS Weight3",
count=results_df$n_detected_in_causal_weight3,
ifcausal=T))
plot_df <- rbind(plot_df,
data.frame(simutag=results_df$simutag,
method="cTWAS Union",
count=results_df$n_detected_combined-results_df$n_detected_in_causal_combined,
ifcausal=F))
plot_df <- rbind(plot_df,
data.frame(simutag=results_df$simutag,
method="cTWAS Union",
count=results_df$n_detected_in_causal_combined,
ifcausal=T))Figure
For the cTWAS analysis, each tissue was analyzed individually and the results were combined.
plot_df$method <- factor(plot_df$method, levels=c("TWAS", "cTWAS G+T", "cTWAS G", "cTWAS G - Ind Var", "cTWAS Union", "cTWAS Weight1", "cTWAS Weight2", "cTWAS Weight3"))
library(ggpubr)
colset = c("#ebebeb", "#fb8072")
ggbarplot(plot_df,
x = "method",
y = "count",
add = "mean_se",
fill = "ifcausal",
legend = "none",
ylab="Count",
xlab="",
palette = colset) + grids(linetype = "dashed") + theme(axis.text.x = element_text(angle = 45, vjust = 1, hjust=1))
plot_df_2 <- plot_df
plot_df_2$method %in% c("TWAS", "cTWAS Union", "cTWAS Weight1", "cTWAS Weight2", "cTWAS Weight3") [1] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE
[13] TRUE TRUE TRUE TRUE TRUE TRUE FALSE FALSE FALSE FALSE FALSE FALSE
[25] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
[37] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUEplot_df_2$ifcausal <- as.character(plot_df_2$ifcausal+1)
plot_df_2$count[plot_df_2$method=="cTWAS G" & plot_df_2$ifcausal=="2"] <- plot_df_2$count[plot_df_2$method=="cTWAS G" & plot_df_2$ifcausal=="2"] - plot_df_2$count[plot_df_2$method=="cTWAS G+T" & plot_df_2$ifcausal=="2"]
plot_df_2 <- rbind(plot_df_2, data.frame(simutag=simutags,
method="cTWAS G",
count=plot_df_2$count[plot_df_2$method=="cTWAS G+T" & plot_df_2$ifcausal=="2"],
ifcausal="3"))
plot_df_3 <- plot_df_2
plot_df_2 <- plot_df_2[plot_df_2$method %in% c("TWAS", "cTWAS Union", "cTWAS Weight1", "cTWAS Weight2", "cTWAS Weight3"),]
plot_df_2$method <- droplevels(plot_df_2$method)
levels(plot_df_2$method) <- c("TWAS", "cTWAS Union", "cTWAS Primary", "cTWAS Secondary", "cTWAS Null")
plot_df_2$ifcausal[plot_df_2$ifcausal=="1"] <- "FP"
plot_df_2$ifcausal[plot_df_2$ifcausal=="2"] <- "TP"
colset = c("#ebebeb", "#fb8072", "#8dd3c7")
ggbarplot(plot_df_2,
x = "method",
y = "count",
add = "mean_se",
fill = "ifcausal",
legend = "right",
ylab="Count",
xlab="",
palette = colset) + grids(linetype = "dashed") + theme(axis.text.x = element_text(angle = 45, vjust = 1, hjust=1)) + guides(fill=guide_legend(title=""))
plot_df_3 <- plot_df_3[plot_df_3$method %in% c("cTWAS G", "cTWAS Union"),]
plot_df_3$method <- droplevels(plot_df_3$method)
levels(plot_df_3$method) <- c("cTWAS Joint", "cTWAS Union")
plot_df_3$method <- relevel(plot_df_3$method, "cTWAS Union")
plot_df_3$ifcausal[plot_df_3$ifcausal=="1"] <- "FP"
plot_df_3$ifcausal[plot_df_3$ifcausal=="2"] <- "TP"
plot_df_3$ifcausal[plot_df_3$ifcausal=="3"] <- "TP and Tissue Identified\nby Joint Analysis"
ggbarplot(plot_df_3,
x = "method",
y = "count",
add = "mean_se",
fill = "ifcausal",
legend = "right",
ylab="Count",
xlab="",
palette = colset) + grids(linetype = "dashed") + theme(axis.text.x = element_text(angle = 45, vjust = 1, hjust=1)) + guides(fill=guide_legend(title=""))
Simulation 3: Three causal tissues with unequal PVE
Separate effect size parameters
For the cTWAS analysis, each tissue had its own prior inclusion parameter and effect size parameter.
results_dir <- "/project2/xinhe/shengqian/cTWAS/cTWAS_simulation/simulation_uncorrelated/"
runtag = "ukb-s80.45-3_uncorr"
configtag <- 2
simutags <- paste(3, 1:3, sep = "-")
thin <- 0.1
sample_size <- 45000
PIP_threshold <- 0.8
results_df <- data.frame(simutag=as.character(),
n_causal=as.integer(),
n_causal_combined=as.integer(),
n_detected_pip=as.integer(),
n_detected_pip_in_causal=as.integer(),
n_detected_comb_pip=as.integer(),
n_detected_comb_pip_in_causal=as.integer(),
pve_snp=as.numeric(),
pve_weight1=as.numeric(),
pve_weight2=as.numeric(),
pve_weight3=as.numeric(),
prior_weight1=as.numeric(),
prior_weight2=as.numeric(),
prior_weight3=as.numeric(),
prior_var_snp=as.numeric(),
prior_var_weight1=as.numeric(),
prior_var_weight2=as.numeric(),
prior_var_weight3=as.numeric())
for (i in 1:length(simutags)){
simutag <- simutags[i]
#load genes with true simulated effect
load(paste0(results_dir, runtag, "_simu", simutag, "-pheno.Rd"))
true_genes <- unlist(sapply(1:22, function(x){phenores$batch[[x]]$id.cgene}))
true_genes_combined <- unique(sapply(true_genes, function(x){unlist(strsplit(x, "[|]"))[1]}))
#load cTWAS results
ctwas_res <- data.table::fread(paste0(results_dir, runtag, "_simu", simutag, "_config", configtag, "_LDR.susieIrss.txt"))
ctwas_gene_res <- ctwas_res[ctwas_res$type!="SNP",]
#number of causal genes
n_causal <- length(true_genes)
n_causal_combined <- length(true_genes_combined)
#number of gene+tissue combinations with cTWAS PIP > threshold
n_ctwas_genes <- sum(ctwas_gene_res$susie_pip > PIP_threshold)
#number of cTWAS genes that are causal
n_causal_detected <- sum(ctwas_gene_res$id[ctwas_gene_res$susie_pip > PIP_threshold] %in% true_genes)
#collapse gene+tissues to genes and compute combined PIP
ctwas_gene_res$gene <- sapply(ctwas_gene_res$id, function(x){unlist(strsplit(x,"[|]"))[1]})
ctwas_gene_res_combined <- aggregate(ctwas_gene_res$susie_pip, by=list(ctwas_gene_res$gene), FUN=sum)
colnames(ctwas_gene_res_combined) <- c("gene", "pip_combined")
#number of genes with combined PIP > threshold
n_ctwas_genes_combined <- sum(ctwas_gene_res_combined$pip_combined > PIP_threshold)
#number of cTWAS genes using combined PIP that are causal
n_causal_detected_combined <- sum(ctwas_gene_res_combined$gene[ctwas_gene_res_combined$pip_combined > PIP_threshold] %in% true_genes_combined)
#collect number of SNPs analyzed by cTWAS
ctwas_res_s1 <- data.table::fread(paste0(results_dir, runtag, "_simu", simutag, "_config", configtag, "_LDR.s1.susieIrss.txt"))
n_snps <- sum(ctwas_res_s1$type=="SNP")/thin
rm(ctwas_res_s1)
#load estimated parameters
load(paste0(results_dir, runtag, "_simu", simutag, "_config", configtag, "_LDR.s2.susieIrssres.Rd"))
#estimated group prior (all iterations)
estimated_group_prior_all <- group_prior_rec
estimated_group_prior_all["SNP",] <- estimated_group_prior_all["SNP",]*thin #adjust parameter to account for thin argument
#estimated group prior variance (all iterations)
estimated_group_prior_var_all <- group_prior_var_rec
#set group size
group_size <- c(table(ctwas_gene_res$type), structure(n_snps, names="SNP"))
group_size <- group_size[rownames(estimated_group_prior_all)]
#estimated group PVE (all iterations)
estimated_group_pve_all <- estimated_group_prior_var_all*estimated_group_prior_all*group_size/sample_size #check PVE calculation
results_current <- data.frame(simutag=as.character(simutag),
n_causal=as.integer(n_causal),
n_causal_combined=as.integer(n_causal_combined),
n_detected_pip=as.integer(n_ctwas_genes),
n_detected_pip_in_causal=as.integer(n_causal_detected),
n_detected_comb_pip=as.integer(n_ctwas_genes_combined),
n_detected_comb_pip_in_causal=as.integer(n_causal_detected_combined),
pve_snp=as.numeric(rev(estimated_group_pve_all["SNP",])[1]),
pve_weight1=as.numeric(rev(estimated_group_pve_all["Liver_harmonized",])[1]),
pve_weight2=as.numeric(rev(estimated_group_pve_all["Lung_harmonized",])[1]),
pve_weight3=as.numeric(rev(estimated_group_pve_all["Brain_Hippocampus_harmonized",])[1]),
prior_snp=as.numeric(rev(estimated_group_prior_var_all["SNP",])[1]),
prior_weight1=as.numeric(rev(estimated_group_prior_all["Liver_harmonized",])[1]),
prior_weight2=as.numeric(rev(estimated_group_prior_all["Lung_harmonized",])[1]),
prior_weight3=as.numeric(rev(estimated_group_prior_all["Brain_Hippocampus_harmonized",])[1]),
prior_var_snp=as.numeric(rev(estimated_group_prior_var_all["SNP",])[1]),
prior_var_weight1=as.numeric(rev(estimated_group_prior_var_all["Liver_harmonized",])[1]),
prior_var_weight2=as.numeric(rev(estimated_group_prior_var_all["Lung_harmonized",])[1]),
prior_var_weight3=as.numeric(rev(estimated_group_prior_var_all["Brain_Hippocampus_harmonized",])[1]))
results_df <- rbind(results_df, results_current)
}
#results using PIP threshold (gene+tissue)
results_df[,c("simutag", "n_causal", "n_detected_pip", "n_detected_pip_in_causal")] simutag n_causal n_detected_pip n_detected_pip_in_causal
1 3-1 129 40 10
2 3-2 154 87 25
3 3-3 119 70 14#mean percent causal using PIP > 0.8
sum(results_df$n_detected_pip_in_causal)/sum(results_df$n_detected_pip)[1] 0.248731#results using combined PIP threshold
results_df[,c("simutag", "n_causal_combined", "n_detected_comb_pip", "n_detected_comb_pip_in_causal")] simutag n_causal_combined n_detected_comb_pip n_detected_comb_pip_in_causal
1 3-1 129 42 15
2 3-2 153 92 36
3 3-3 118 70 15#mean percent causal using combined PIP > 0.8
sum(results_df$n_detected_comb_pip_in_causal)/sum(results_df$n_detected_comb_pip)[1] 0.3235294#prior inclusion and mean prior inclusion
results_df[,c(which(colnames(results_df)=="simutag"), setdiff(grep("prior", names(results_df)), grep("prior_var", names(results_df))))] simutag prior_snp prior_weight1 prior_weight2 prior_weight3
1 3-1 8.238626 0.008787670 0.0006807431 0.002515466
2 3-2 7.233992 0.004264379 0.0030664678 0.010199889
3 3-3 8.058730 0.006422694 0.0041318938 0.003979957colMeans(results_df[,setdiff(grep("prior", names(results_df)), grep("prior_var", names(results_df)))]) prior_snp prior_weight1 prior_weight2 prior_weight3
7.843782664 0.006491581 0.002626368 0.005565104 #prior variance and mean prior variance
results_df[,c(which(colnames(results_df)=="simutag"), grep("prior_var", names(results_df)))] simutag prior_var_snp prior_var_weight1 prior_var_weight2 prior_var_weight3
1 3-1 8.238626 16.39297 5.674244 9.703997
2 3-2 7.233992 27.83989 17.045111 13.232442
3 3-3 8.058730 32.65081 6.414417 14.332842colMeans(results_df[,grep("prior_var", names(results_df))]) prior_var_snp prior_var_weight1 prior_var_weight2 prior_var_weight3
7.843783 25.627891 9.711257 12.423093 #PVE and mean PVE
results_df[,c(which(colnames(results_df)=="simutag"), grep("pve", names(results_df)))] simutag pve_snp pve_weight1 pve_weight2 pve_weight3
1 3-1 0.2683736 0.02365079 0.0008360605 0.004312446
2 3-2 0.2577457 0.01949116 0.0113131798 0.023844600
3 3-3 0.2616802 0.03442909 0.0057365766 0.010077791colMeans(results_df[,grep("pve", names(results_df))]) pve_snp pve_weight1 pve_weight2 pve_weight3
0.262599859 0.025857015 0.005961939 0.012744945 #store results for figure
plot_df <- rbind(plot_df,
data.frame(simutag=results_df$simutag,
method="cTWAS G - Ind Var",
count=results_df$n_detected_comb_pip-results_df$n_detected_comb_pip_in_causal,
ifcausal=F))
plot_df <- rbind(plot_df,
data.frame(simutag=results_df$simutag,
method="cTWAS G - Ind Var",
count=results_df$n_detected_comb_pip_in_causal,
ifcausal=T))Individual tissue analyses
For the cTWAS analysis, each tissue was analyzed individually and the results were combined.
results_dir <- "/project2/xinhe/shengqian/cTWAS/cTWAS_simulation/simulation_uncorrelated/"
runtag = "ukb-s80.45-3_uncorr"
configtag <- 2
simutags <- paste(3, 1:3, sep = "-")
thin <- 0.1
sample_size <- 45000
PIP_threshold <- 0.8
results_df <- data.frame(simutag=as.character(),
n_causal_combined=as.integer(),
n_detected_weight1=as.integer(),
n_detected_in_causal_weight1=as.integer(),
n_detected_weight2=as.integer(),
n_detected_in_causal_weight2=as.integer(),
n_detected_weight3=as.integer(),
n_detected_in_causal_weight3=as.integer(),
n_detected_combined=as.integer(),
n_detected_in_causal_combined=as.integer())
for (i in 1:length(simutags)){
simutag <- simutags[i]
#load genes with true simulated effect
load(paste0(results_dir, runtag, "_simu", simutag, "-pheno.Rd"))
true_genes <- unlist(sapply(1:22, function(x){phenores$batch[[x]]$id.cgene}))
true_genes_combined <- unique(sapply(true_genes, function(x){unlist(strsplit(x, "[|]"))[1]}))
#number of causal genes
n_causal_combined <- length(true_genes_combined)
#load cTWAS results for weight1
ctwas_res <- data.table::fread(paste0(results_dir, runtag, "_simu", simutag, "_config", configtag, "_LDR_weight1.susieIrss.txt"))
ctwas_gene_res <- ctwas_res[ctwas_res$type!="SNP",]
#number of genes with cTWAS PIP > threshold in weight1
ctwas_genes_weight1 <- ctwas_gene_res$id[ctwas_gene_res$susie_pip > PIP_threshold]
n_ctwas_genes_weight1 <- length(ctwas_genes_weight1)
n_causal_detected_weight1 <- sum(ctwas_genes_weight1 %in% true_genes_combined)
#load cTWAS results for weight2
ctwas_res <- data.table::fread(paste0(results_dir, runtag, "_simu", simutag, "_config", configtag, "_LDR_weight2.susieIrss.txt"))
ctwas_gene_res <- ctwas_res[ctwas_res$type!="SNP",]
#number of genes with cTWAS PIP > threshold in weight1
ctwas_genes_weight2 <- ctwas_gene_res$id[ctwas_gene_res$susie_pip > PIP_threshold]
n_ctwas_genes_weight2 <- length(ctwas_genes_weight2)
n_causal_detected_weight2 <- sum(ctwas_genes_weight2 %in% true_genes_combined)
#load cTWAS results for weight3
ctwas_res <- data.table::fread(paste0(results_dir, runtag, "_simu", simutag, "_config", configtag, "_LDR_weight3.susieIrss.txt"))
ctwas_gene_res <- ctwas_res[ctwas_res$type!="SNP",]
#number of genes with cTWAS PIP > threshold in weight3
ctwas_genes_weight3 <- ctwas_gene_res$id[ctwas_gene_res$susie_pip > PIP_threshold]
n_ctwas_genes_weight3 <- length(ctwas_genes_weight3)
n_causal_detected_weight3 <- sum(ctwas_genes_weight3 %in% true_genes_combined)
#combined analysis
ctwas_genes_combined <- unique(c(ctwas_genes_weight1, ctwas_genes_weight2, ctwas_genes_weight3))
n_ctwas_genes_combined <- length(ctwas_genes_combined)
n_causal_detected_combined <- sum(ctwas_genes_combined %in% true_genes_combined)
results_current <- data.frame(simutag=as.character(simutag),
n_causal_combined=as.integer(n_causal_combined),
n_detected_weight1=as.integer(n_ctwas_genes_weight1),
n_detected_in_causal_weight1=as.integer(n_causal_detected_weight1),
n_detected_weight2=as.integer(n_ctwas_genes_weight2),
n_detected_in_causal_weight2=as.integer(n_causal_detected_weight2),
n_detected_weight3=as.integer(n_ctwas_genes_weight3),
n_detected_in_causal_weight3=as.integer(n_causal_detected_weight3),
n_detected_combined=as.integer(n_ctwas_genes_combined),
n_detected_in_causal_combined=as.integer(n_causal_detected_combined))
results_df <- rbind(results_df, results_current)
}
#results using weight1
results_df[,c("simutag", colnames(results_df)[grep("weight1", colnames(results_df))])] simutag n_detected_weight1 n_detected_in_causal_weight1
1 3-1 35 12
2 3-2 50 21
3 3-3 45 11#mean percent causal using PIP > 0.8 for weight1
sum(results_df$n_detected_in_causal_weight1)/sum(results_df$n_detected_weight1)[1] 0.3384615#results using weight2
results_df[,c("simutag", colnames(results_df)[grep("weight2", colnames(results_df))])] simutag n_detected_weight2 n_detected_in_causal_weight2
1 3-1 30 12
2 3-2 41 16
3 3-3 18 3#mean percent causal using PIP > 0.8 for weight1
sum(results_df$n_detected_in_causal_weight2)/sum(results_df$n_detected_weight2)[1] 0.3483146#results using weight3
results_df[,c("simutag", colnames(results_df)[grep("weight3", colnames(results_df))])] simutag n_detected_weight3 n_detected_in_causal_weight3
1 3-1 30 10
2 3-2 39 12
3 3-3 40 4#mean percent causal using PIP > 0.8 for weight3
sum(results_df$n_detected_in_causal_weight3)/sum(results_df$n_detected_weight3)[1] 0.2385321#results using combined analysis
results_df[,c("simutag", colnames(results_df)[grep("combined", colnames(results_df))])] simutag n_causal_combined n_detected_combined n_detected_in_causal_combined
1 3-1 129 71 24
2 3-2 153 92 31
3 3-3 118 80 16#mean percent causal using PIP > 0.8 for weight3
sum(results_df$n_detected_in_causal_combined)/sum(results_df$n_detected_combined)[1] 0.2921811#store results for figure
plot_df <- rbind(plot_df,
data.frame(simutag=results_df$simutag,
method="cTWAS Weight1",
count=results_df$n_detected_weight1-results_df$n_detected_in_causal_weight1,
ifcausal=F))
plot_df <- rbind(plot_df,
data.frame(simutag=results_df$simutag,
method="cTWAS Weight1",
count=results_df$n_detected_in_causal_weight1,
ifcausal=T))
plot_df <- rbind(plot_df,
data.frame(simutag=results_df$simutag,
method="cTWAS Weight2",
count=results_df$n_detected_weight2-results_df$n_detected_in_causal_weight2,
ifcausal=F))
plot_df <- rbind(plot_df,
data.frame(simutag=results_df$simutag,
method="cTWAS Weight2",
count=results_df$n_detected_in_causal_weight2,
ifcausal=T))
plot_df <- rbind(plot_df,
data.frame(simutag=results_df$simutag,
method="cTWAS Weight3",
count=results_df$n_detected_weight3-results_df$n_detected_in_causal_weight3,
ifcausal=F))
plot_df <- rbind(plot_df,
data.frame(simutag=results_df$simutag,
method="cTWAS Weight3",
count=results_df$n_detected_in_causal_weight3,
ifcausal=T))
plot_df <- rbind(plot_df,
data.frame(simutag=results_df$simutag,
method="cTWAS Union",
count=results_df$n_detected_combined-results_df$n_detected_in_causal_combined,
ifcausal=F))
plot_df <- rbind(plot_df,
data.frame(simutag=results_df$simutag,
method="cTWAS Union",
count=results_df$n_detected_in_causal_combined,
ifcausal=T))Figure
For the cTWAS analysis, each tissue was analyzed individually and the results were combined.
plot_df$method <- factor(plot_df$method, levels=c("TWAS", "cTWAS G+T", "cTWAS G", "cTWAS G - Ind Var", "cTWAS Union", "cTWAS Weight1", "cTWAS Weight2", "cTWAS Weight3"))
library(ggpubr)
colset = c("#ebebeb", "#fb8072")
ggbarplot(plot_df,
x = "method",
y = "count",
add = "mean_se",
fill = "ifcausal",
legend = "none",
ylab="Count",
xlab="",
palette = colset) + grids(linetype = "dashed") + theme(axis.text.x = element_text(angle = 45, vjust = 1, hjust=1))
plot_df_2 <- plot_df
plot_df_2$method %in% c("TWAS", "cTWAS Union", "cTWAS Weight1", "cTWAS Weight2", "cTWAS Weight3") [1] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE
[13] TRUE TRUE TRUE TRUE TRUE TRUE FALSE FALSE FALSE FALSE FALSE FALSE
[25] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
[37] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUEplot_df_2$ifcausal <- as.character(plot_df_2$ifcausal+1)
plot_df_2$count[plot_df_2$method=="cTWAS G" & plot_df_2$ifcausal=="2"] <- plot_df_2$count[plot_df_2$method=="cTWAS G" & plot_df_2$ifcausal=="2"] - plot_df_2$count[plot_df_2$method=="cTWAS G+T" & plot_df_2$ifcausal=="2"]
plot_df_2 <- rbind(plot_df_2, data.frame(simutag=simutags,
method="cTWAS G",
count=plot_df_2$count[plot_df_2$method=="cTWAS G+T" & plot_df_2$ifcausal=="2"],
ifcausal="3"))
plot_df_3 <- plot_df_2
plot_df_2 <- plot_df_2[plot_df_2$method %in% c("TWAS", "cTWAS Union", "cTWAS Weight1", "cTWAS Weight2", "cTWAS Weight3"),]
plot_df_2$method <- droplevels(plot_df_2$method)
levels(plot_df_2$method) <- c("TWAS", "cTWAS Union", "cTWAS Primary", "cTWAS Secondary", "cTWAS Null")
plot_df_2$ifcausal[plot_df_2$ifcausal=="1"] <- "FP"
plot_df_2$ifcausal[plot_df_2$ifcausal=="2"] <- "TP"
colset = c("#ebebeb", "#fb8072", "#8dd3c7")
ggbarplot(plot_df_2,
x = "method",
y = "count",
add = "mean_se",
fill = "ifcausal",
legend = "right",
ylab="Count",
xlab="",
palette = colset) + grids(linetype = "dashed") + theme(axis.text.x = element_text(angle = 45, vjust = 1, hjust=1)) + guides(fill=guide_legend(title=""))
plot_df_3 <- plot_df_3[plot_df_3$method %in% c("cTWAS G", "cTWAS Union"),]
plot_df_3$method <- droplevels(plot_df_3$method)
levels(plot_df_3$method) <- c("cTWAS Joint", "cTWAS Union")
plot_df_3$method <- relevel(plot_df_3$method, "cTWAS Union")
plot_df_3$ifcausal[plot_df_3$ifcausal=="1"] <- "FP"
plot_df_3$ifcausal[plot_df_3$ifcausal=="2"] <- "TP"
plot_df_3$ifcausal[plot_df_3$ifcausal=="3"] <- "TP and Tissue Identified\nby Joint Analysis"
ggbarplot(plot_df_3,
x = "method",
y = "count",
add = "mean_se",
fill = "ifcausal",
legend = "right",
ylab="Count",
xlab="",
palette = colset) + grids(linetype = "dashed") + theme(axis.text.x = element_text(angle = 45, vjust = 1, hjust=1)) + guides(fill=guide_legend(title=""))
Simulation 4: Two causal tissues with unequal PVE
Separate effect size parameters
For the cTWAS analysis, each tissue had its own prior inclusion parameter end effect size parameter.
results_dir <- "/project2/xinhe/shengqian/cTWAS/cTWAS_simulation/simulation_uncorrelated/"
runtag = "ukb-s80.45-3_uncorr"
configtag <- 2
simutags <- paste(4, 1:3, sep = "-")
thin <- 0.1
sample_size <- 45000
PIP_threshold <- 0.8
results_df <- data.frame(simutag=as.character(),
n_causal=as.integer(),
n_causal_combined=as.integer(),
n_detected_pip=as.integer(),
n_detected_pip_in_causal=as.integer(),
n_detected_comb_pip=as.integer(),
n_detected_comb_pip_in_causal=as.integer(),
pve_snp=as.numeric(),
pve_weight1=as.numeric(),
pve_weight2=as.numeric(),
pve_weight3=as.numeric(),
prior_weight1=as.numeric(),
prior_weight2=as.numeric(),
prior_weight3=as.numeric(),
prior_var_snp=as.numeric(),
prior_var_weight1=as.numeric(),
prior_var_weight2=as.numeric(),
prior_var_weight3=as.numeric())
for (i in 1:length(simutags)){
simutag <- simutags[i]
#load genes with true simulated effect
load(paste0(results_dir, runtag, "_simu", simutag, "-pheno.Rd"))
true_genes <- unlist(sapply(1:22, function(x){phenores$batch[[x]]$id.cgene}))
true_genes_combined <- unique(sapply(true_genes, function(x){unlist(strsplit(x, "[|]"))[1]}))
#load cTWAS results
ctwas_res <- data.table::fread(paste0(results_dir, runtag, "_simu", simutag, "_config", configtag, "_LDR.susieIrss.txt"))
ctwas_gene_res <- ctwas_res[ctwas_res$type!="SNP",]
#number of causal genes
n_causal <- length(true_genes)
n_causal_combined <- length(true_genes_combined)
#number of gene+tissue combinations with cTWAS PIP > threshold
n_ctwas_genes <- sum(ctwas_gene_res$susie_pip > PIP_threshold)
#number of cTWAS genes that are causal
n_causal_detected <- sum(ctwas_gene_res$id[ctwas_gene_res$susie_pip > PIP_threshold] %in% true_genes)
#collapse gene+tissues to genes and compute combined PIP
ctwas_gene_res$gene <- sapply(ctwas_gene_res$id, function(x){unlist(strsplit(x,"[|]"))[1]})
ctwas_gene_res_combined <- aggregate(ctwas_gene_res$susie_pip, by=list(ctwas_gene_res$gene), FUN=sum)
colnames(ctwas_gene_res_combined) <- c("gene", "pip_combined")
#number of genes with combined PIP > threshold
n_ctwas_genes_combined <- sum(ctwas_gene_res_combined$pip_combined > PIP_threshold)
#number of cTWAS genes using combined PIP that are causal
n_causal_detected_combined <- sum(ctwas_gene_res_combined$gene[ctwas_gene_res_combined$pip_combined > PIP_threshold] %in% true_genes_combined)
#collect number of SNPs analyzed by cTWAS
ctwas_res_s1 <- data.table::fread(paste0(results_dir, runtag, "_simu", simutag, "_config", configtag, "_LDR.s1.susieIrss.txt"))
n_snps <- sum(ctwas_res_s1$type=="SNP")/thin
rm(ctwas_res_s1)
#load estimated parameters
load(paste0(results_dir, runtag, "_simu", simutag, "_config", configtag, "_LDR.s2.susieIrssres.Rd"))
#estimated group prior (all iterations)
estimated_group_prior_all <- group_prior_rec
estimated_group_prior_all["SNP",] <- estimated_group_prior_all["SNP",]*thin #adjust parameter to account for thin argument
#estimated group prior variance (all iterations)
estimated_group_prior_var_all <- group_prior_var_rec
#set group size
group_size <- c(table(ctwas_gene_res$type), structure(n_snps, names="SNP"))
group_size <- group_size[rownames(estimated_group_prior_all)]
#estimated group PVE (all iterations)
estimated_group_pve_all <- estimated_group_prior_var_all*estimated_group_prior_all*group_size/sample_size #check PVE calculation
results_current <- data.frame(simutag=as.character(simutag),
n_causal=as.integer(n_causal),
n_causal_combined=as.integer(n_causal_combined),
n_detected_pip=as.integer(n_ctwas_genes),
n_detected_pip_in_causal=as.integer(n_causal_detected),
n_detected_comb_pip=as.integer(n_ctwas_genes_combined),
n_detected_comb_pip_in_causal=as.integer(n_causal_detected_combined),
pve_snp=as.numeric(rev(estimated_group_pve_all["SNP",])[1]),
pve_weight1=as.numeric(rev(estimated_group_pve_all["Liver_harmonized",])[1]),
pve_weight2=as.numeric(rev(estimated_group_pve_all["Lung_harmonized",])[1]),
pve_weight3=as.numeric(rev(estimated_group_pve_all["Brain_Hippocampus_harmonized",])[1]),
prior_snp=as.numeric(rev(estimated_group_prior_var_all["SNP",])[1]),
prior_weight1=as.numeric(rev(estimated_group_prior_all["Liver_harmonized",])[1]),
prior_weight2=as.numeric(rev(estimated_group_prior_all["Lung_harmonized",])[1]),
prior_weight3=as.numeric(rev(estimated_group_prior_all["Brain_Hippocampus_harmonized",])[1]),
prior_var_snp=as.numeric(rev(estimated_group_prior_var_all["SNP",])[1]),
prior_var_weight1=as.numeric(rev(estimated_group_prior_var_all["Liver_harmonized",])[1]),
prior_var_weight2=as.numeric(rev(estimated_group_prior_var_all["Lung_harmonized",])[1]),
prior_var_weight3=as.numeric(rev(estimated_group_prior_var_all["Brain_Hippocampus_harmonized",])[1]))
results_df <- rbind(results_df, results_current)
}
#results using PIP threshold (gene+tissue)
results_df[,c("simutag", "n_causal", "n_detected_pip", "n_detected_pip_in_causal")] simutag n_causal n_detected_pip n_detected_pip_in_causal
1 4-1 85 63 11
2 4-2 93 68 8
3 4-3 116 37 20#mean percent causal using PIP > 0.8
sum(results_df$n_detected_pip_in_causal)/sum(results_df$n_detected_pip)[1] 0.2321429#results using combined PIP threshold
results_df[,c("simutag", "n_causal_combined", "n_detected_comb_pip", "n_detected_comb_pip_in_causal")] simutag n_causal_combined n_detected_comb_pip n_detected_comb_pip_in_causal
1 4-1 85 68 17
2 4-2 93 73 22
3 4-3 116 39 23#mean percent causal using combined PIP > 0.8
sum(results_df$n_detected_comb_pip_in_causal)/sum(results_df$n_detected_comb_pip)[1] 0.3444444#prior inclusion and mean prior inclusion
results_df[,c(which(colnames(results_df)=="simutag"), setdiff(grep("prior", names(results_df)), grep("prior_var", names(results_df))))] simutag prior_snp prior_weight1 prior_weight2 prior_weight3
1 4-1 7.302342 0.006883607 0.002361883 0.004268799
2 4-2 8.056925 0.004889306 0.001860718 0.002974719
3 4-3 7.486974 0.012656143 0.008062868 0.001902555colMeans(results_df[,setdiff(grep("prior", names(results_df)), grep("prior_var", names(results_df)))]) prior_snp prior_weight1 prior_weight2 prior_weight3
7.615413561 0.008143019 0.004095156 0.003048691 #prior variance and mean prior variance
results_df[,c(which(colnames(results_df)=="simutag"), grep("prior_var", names(results_df)))] simutag prior_var_snp prior_var_weight1 prior_var_weight2 prior_var_weight3
1 4-1 7.302342 19.97816 18.613663 9.297576
2 4-2 8.056925 23.92186 24.515628 40.806612
3 4-3 7.486974 20.23736 6.903486 1.958156colMeans(results_df[,grep("prior_var", names(results_df))]) prior_var_snp prior_var_weight1 prior_var_weight2 prior_var_weight3
7.615414 21.379124 16.677592 17.354114 #PVE and mean PVE
results_df[,c(which(colnames(results_df)=="simutag"), grep("pve", names(results_df)))] simutag pve_snp pve_weight1 pve_weight2 pve_weight3
1 4-1 0.2574665 0.02257802 0.009515612 0.0070118080
2 4-2 0.2477735 0.01920244 0.009873477 0.0214452526
3 4-3 0.2559470 0.04205035 0.012047707 0.0006581717colMeans(results_df[,grep("pve", names(results_df))]) pve_snp pve_weight1 pve_weight2 pve_weight3
0.253729003 0.027943604 0.010478932 0.009705077 #store results for figure
plot_df <- rbind(plot_df,
data.frame(simutag=results_df$simutag,
method="cTWAS G - Ind Var",
count=results_df$n_detected_comb_pip-results_df$n_detected_comb_pip_in_causal,
ifcausal=F))
plot_df <- rbind(plot_df,
data.frame(simutag=results_df$simutag,
method="cTWAS G - Ind Var",
count=results_df$n_detected_comb_pip_in_causal,
ifcausal=T))Individual tissue analyses
For the cTWAS analysis, each tissue was analyzed individually and the results were combined.
results_dir <- "/project2/xinhe/shengqian/cTWAS/cTWAS_simulation/simulation_uncorrelated/"
runtag = "ukb-s80.45-3_uncorr"
configtag <- 2
simutags <- paste(4, 1:3, sep = "-")
thin <- 0.1
sample_size <- 45000
PIP_threshold <- 0.8
results_df <- data.frame(simutag=as.character(),
n_causal_combined=as.integer(),
n_detected_weight1=as.integer(),
n_detected_in_causal_weight1=as.integer(),
n_detected_weight2=as.integer(),
n_detected_in_causal_weight2=as.integer(),
n_detected_weight3=as.integer(),
n_detected_in_causal_weight3=as.integer(),
n_detected_combined=as.integer(),
n_detected_in_causal_combined=as.integer())
for (i in 1:length(simutags)){
simutag <- simutags[i]
#load genes with true simulated effect
load(paste0(results_dir, runtag, "_simu", simutag, "-pheno.Rd"))
true_genes <- unlist(sapply(1:22, function(x){phenores$batch[[x]]$id.cgene}))
true_genes_combined <- unique(sapply(true_genes, function(x){unlist(strsplit(x, "[|]"))[1]}))
#number of causal genes
n_causal_combined <- length(true_genes_combined)
#load cTWAS results for weight1
ctwas_res <- data.table::fread(paste0(results_dir, runtag, "_simu", simutag, "_config", configtag, "_LDR_weight1.susieIrss.txt"))
ctwas_gene_res <- ctwas_res[ctwas_res$type!="SNP",]
#number of genes with cTWAS PIP > threshold in weight1
ctwas_genes_weight1 <- ctwas_gene_res$id[ctwas_gene_res$susie_pip > PIP_threshold]
n_ctwas_genes_weight1 <- length(ctwas_genes_weight1)
n_causal_detected_weight1 <- sum(ctwas_genes_weight1 %in% true_genes_combined)
#load cTWAS results for weight2
ctwas_res <- data.table::fread(paste0(results_dir, runtag, "_simu", simutag, "_config", configtag, "_LDR_weight2.susieIrss.txt"))
ctwas_gene_res <- ctwas_res[ctwas_res$type!="SNP",]
#number of genes with cTWAS PIP > threshold in weight1
ctwas_genes_weight2 <- ctwas_gene_res$id[ctwas_gene_res$susie_pip > PIP_threshold]
n_ctwas_genes_weight2 <- length(ctwas_genes_weight2)
n_causal_detected_weight2 <- sum(ctwas_genes_weight2 %in% true_genes_combined)
#load cTWAS results for weight3
ctwas_res <- data.table::fread(paste0(results_dir, runtag, "_simu", simutag, "_config", configtag, "_LDR_weight3.susieIrss.txt"))
ctwas_gene_res <- ctwas_res[ctwas_res$type!="SNP",]
#number of genes with cTWAS PIP > threshold in weight3
ctwas_genes_weight3 <- ctwas_gene_res$id[ctwas_gene_res$susie_pip > PIP_threshold]
n_ctwas_genes_weight3 <- length(ctwas_genes_weight3)
n_causal_detected_weight3 <- sum(ctwas_genes_weight3 %in% true_genes_combined)
#combined analysis
ctwas_genes_combined <- unique(c(ctwas_genes_weight1, ctwas_genes_weight2, ctwas_genes_weight3))
n_ctwas_genes_combined <- length(ctwas_genes_combined)
n_causal_detected_combined <- sum(ctwas_genes_combined %in% true_genes_combined)
results_current <- data.frame(simutag=as.character(simutag),
n_causal_combined=as.integer(n_causal_combined),
n_detected_weight1=as.integer(n_ctwas_genes_weight1),
n_detected_in_causal_weight1=as.integer(n_causal_detected_weight1),
n_detected_weight2=as.integer(n_ctwas_genes_weight2),
n_detected_in_causal_weight2=as.integer(n_causal_detected_weight2),
n_detected_weight3=as.integer(n_ctwas_genes_weight3),
n_detected_in_causal_weight3=as.integer(n_causal_detected_weight3),
n_detected_combined=as.integer(n_ctwas_genes_combined),
n_detected_in_causal_combined=as.integer(n_causal_detected_combined))
results_df <- rbind(results_df, results_current)
}
#results using weight1
results_df[,c("simutag", colnames(results_df)[grep("weight1", colnames(results_df))])] simutag n_detected_weight1 n_detected_in_causal_weight1
1 4-1 47 14
2 4-2 36 16
3 4-3 29 16#mean percent causal using PIP > 0.8 for weight1
sum(results_df$n_detected_in_causal_weight1)/sum(results_df$n_detected_weight1)[1] 0.4107143#results using weight2
results_df[,c("simutag", colnames(results_df)[grep("weight2", colnames(results_df))])] simutag n_detected_weight2 n_detected_in_causal_weight2
1 4-1 25 5
2 4-2 31 9
3 4-3 21 9#mean percent causal using PIP > 0.8 for weight1
sum(results_df$n_detected_in_causal_weight2)/sum(results_df$n_detected_weight2)[1] 0.2987013#results using weight3
results_df[,c("simutag", colnames(results_df)[grep("weight3", colnames(results_df))])] simutag n_detected_weight3 n_detected_in_causal_weight3
1 4-1 31 5
2 4-2 36 10
3 4-3 28 6#mean percent causal using PIP > 0.8 for weight3
sum(results_df$n_detected_in_causal_weight3)/sum(results_df$n_detected_weight3)[1] 0.2210526#results using combined analysis
results_df[,c("simutag", colnames(results_df)[grep("combined", colnames(results_df))])] simutag n_causal_combined n_detected_combined n_detected_in_causal_combined
1 4-1 85 79 17
2 4-2 93 79 23
3 4-3 116 61 22#mean percent causal using PIP > 0.8 for weight3
sum(results_df$n_detected_in_causal_combined)/sum(results_df$n_detected_combined)[1] 0.283105#store results for figure
plot_df <- rbind(plot_df,
data.frame(simutag=results_df$simutag,
method="cTWAS Weight1",
count=results_df$n_detected_weight1-results_df$n_detected_in_causal_weight1,
ifcausal=F))
plot_df <- rbind(plot_df,
data.frame(simutag=results_df$simutag,
method="cTWAS Weight1",
count=results_df$n_detected_in_causal_weight1,
ifcausal=T))
plot_df <- rbind(plot_df,
data.frame(simutag=results_df$simutag,
method="cTWAS Weight2",
count=results_df$n_detected_weight2-results_df$n_detected_in_causal_weight2,
ifcausal=F))
plot_df <- rbind(plot_df,
data.frame(simutag=results_df$simutag,
method="cTWAS Weight2",
count=results_df$n_detected_in_causal_weight2,
ifcausal=T))
plot_df <- rbind(plot_df,
data.frame(simutag=results_df$simutag,
method="cTWAS Weight3",
count=results_df$n_detected_weight3-results_df$n_detected_in_causal_weight3,
ifcausal=F))
plot_df <- rbind(plot_df,
data.frame(simutag=results_df$simutag,
method="cTWAS Weight3",
count=results_df$n_detected_in_causal_weight3,
ifcausal=T))
plot_df <- rbind(plot_df,
data.frame(simutag=results_df$simutag,
method="cTWAS Union",
count=results_df$n_detected_combined-results_df$n_detected_in_causal_combined,
ifcausal=F))
plot_df <- rbind(plot_df,
data.frame(simutag=results_df$simutag,
method="cTWAS Union",
count=results_df$n_detected_in_causal_combined,
ifcausal=T))Figure
For the cTWAS analysis, each tissue was analyzed individually and the results were combined.
plot_df$method <- factor(plot_df$method, levels=c("TWAS", "cTWAS G+T", "cTWAS G", "cTWAS G - Ind Var", "cTWAS Union", "cTWAS Weight1", "cTWAS Weight2", "cTWAS Weight3"))
library(ggpubr)
colset = c("#ebebeb", "#fb8072")
ggbarplot(plot_df,
x = "method",
y = "count",
add = "mean_se",
fill = "ifcausal",
legend = "none",
ylab="Count",
xlab="",
palette = colset) + grids(linetype = "dashed") + theme(axis.text.x = element_text(angle = 45, vjust = 1, hjust=1))
plot_df_2 <- plot_df
plot_df_2$method %in% c("TWAS", "cTWAS Union", "cTWAS Weight1", "cTWAS Weight2", "cTWAS Weight3") [1] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE
[13] TRUE TRUE TRUE TRUE TRUE TRUE FALSE FALSE FALSE FALSE FALSE FALSE
[25] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
[37] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUEplot_df_2$ifcausal <- as.character(plot_df_2$ifcausal+1)
plot_df_2$count[plot_df_2$method=="cTWAS G" & plot_df_2$ifcausal=="2"] <- plot_df_2$count[plot_df_2$method=="cTWAS G" & plot_df_2$ifcausal=="2"] - plot_df_2$count[plot_df_2$method=="cTWAS G+T" & plot_df_2$ifcausal=="2"]
plot_df_2 <- rbind(plot_df_2, data.frame(simutag=simutags,
method="cTWAS G",
count=plot_df_2$count[plot_df_2$method=="cTWAS G+T" & plot_df_2$ifcausal=="2"],
ifcausal="3"))
plot_df_3 <- plot_df_2
plot_df_2 <- plot_df_2[plot_df_2$method %in% c("TWAS", "cTWAS Union", "cTWAS Weight1", "cTWAS Weight2", "cTWAS Weight3"),]
plot_df_2$method <- droplevels(plot_df_2$method)
levels(plot_df_2$method) <- c("TWAS", "cTWAS Union", "cTWAS Primary", "cTWAS Secondary", "cTWAS Null")
plot_df_2$ifcausal[plot_df_2$ifcausal=="1"] <- "FP"
plot_df_2$ifcausal[plot_df_2$ifcausal=="2"] <- "TP"
colset = c("#ebebeb", "#fb8072", "#8dd3c7")
ggbarplot(plot_df_2,
x = "method",
y = "count",
add = "mean_se",
fill = "ifcausal",
legend = "right",
ylab="Count",
xlab="",
palette = colset) + grids(linetype = "dashed") + theme(axis.text.x = element_text(angle = 45, vjust = 1, hjust=1)) + guides(fill=guide_legend(title=""))
plot_df_3 <- plot_df_3[plot_df_3$method %in% c("cTWAS G", "cTWAS Union"),]
plot_df_3$method <- droplevels(plot_df_3$method)
levels(plot_df_3$method) <- c("cTWAS Joint", "cTWAS Union")
plot_df_3$method <- relevel(plot_df_3$method, "cTWAS Union")
plot_df_3$ifcausal[plot_df_3$ifcausal=="1"] <- "FP"
plot_df_3$ifcausal[plot_df_3$ifcausal=="2"] <- "TP"
plot_df_3$ifcausal[plot_df_3$ifcausal=="3"] <- "TP and Tissue Identified\nby Joint Analysis"
ggbarplot(plot_df_3,
x = "method",
y = "count",
add = "mean_se",
fill = "ifcausal",
legend = "right",
ylab="Count",
xlab="",
palette = colset) + grids(linetype = "dashed") + theme(axis.text.x = element_text(angle = 45, vjust = 1, hjust=1)) + guides(fill=guide_legend(title=""))
sessionInfo()R version 4.1.0 (2021-05-18)
Platform: x86_64-pc-linux-gnu (64-bit)
Running under: CentOS Linux 7 (Core)
Matrix products: default
BLAS/LAPACK: /software/openblas-0.3.13-el7-x86_64/lib/libopenblas_haswellp-r0.3.13.so
locale:
[1] LC_CTYPE=en_US.UTF-8 LC_NUMERIC=C
[3] LC_TIME=en_US.UTF-8 LC_COLLATE=en_US.UTF-8
[5] LC_MONETARY=en_US.UTF-8 LC_MESSAGES=en_US.UTF-8
[7] LC_PAPER=en_US.UTF-8 LC_NAME=C
[9] LC_ADDRESS=C LC_TELEPHONE=C
[11] LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C
attached base packages:
[1] stats graphics grDevices utils datasets methods base
other attached packages:
[1] ggpubr_0.6.0 ggplot2_3.4.0 workflowr_1.7.0
loaded via a namespace (and not attached):
[1] tidyselect_1.2.0 xfun_0.35 bslib_0.4.1 purrr_1.0.2
[5] carData_3.0-4 colorspace_2.0-3 vctrs_0.6.3 generics_0.1.3
[9] htmltools_0.5.4 yaml_2.3.6 utf8_1.2.2 rlang_1.1.1
[13] jquerylib_0.1.4 later_1.3.0 pillar_1.8.1 glue_1.6.2
[17] withr_2.5.0 DBI_1.1.3 lifecycle_1.0.3 stringr_1.5.0
[21] ggsignif_0.6.4 munsell_0.5.0 gtable_0.3.1 evaluate_0.19
[25] labeling_0.4.2 knitr_1.41 callr_3.7.3 fastmap_1.1.0
[29] httpuv_1.6.7 ps_1.7.2 fansi_1.0.3 highr_0.9
[33] broom_1.0.2 Rcpp_1.0.9 backports_1.2.1 promises_1.2.0.1
[37] scales_1.2.1 cachem_1.0.6 jsonlite_1.8.4 abind_1.4-5
[41] farver_2.1.0 fs_1.5.2 digest_0.6.31 stringi_1.7.8
[45] rstatix_0.7.2 processx_3.8.0 dplyr_1.0.10 getPass_0.2-2
[49] rprojroot_2.0.3 grid_4.1.0 cli_3.6.1 tools_4.1.0
[53] magrittr_2.0.3 sass_0.4.4 tibble_3.1.8 car_3.1-1
[57] tidyr_1.3.0 whisker_0.4.1 pkgconfig_2.0.3 data.table_1.14.6
[61] assertthat_0.2.1 rmarkdown_2.19 httr_1.4.4 rstudioapi_0.14
[65] R6_2.5.1 git2r_0.30.1 compiler_4.1.0