In CreRecombinase/RColumbo: R Package to accompany the Columbo method

In this notebook we'll compare the results of elasticnet, varbvs and GCTA at explaining expression heritability

We'll start with the VB data

Now the Elasticnet data

library(rhdf5)
library(ggplot2)
library(dplyr)
el_files <- dir("/media/nwknoblauch/Data/GTEx/cis_elasticnet/",full.names = T)
h5ls(el_files[1])
overall <- bind_rows(lapply(el_files,read_h5_df,groupname="overall"))
lambda_min <- bind_rows(lapply(el_files,read_h5_df,groupname="lambda_min"))
overall <- group_by(overall,fgeneid) %>% mutate(scaled_estimate=(estimate-mean(estimate))/var(estimate),
                                                scaled_high=(conf.high-mean(conf.high))/var(conf.high),
                                                scaled_low=(conf.low-mean(conf.low))/var(conf.low)) %>%

filter(overall,estimate==max(abs(estimate)))

GCTA results

gcta_estf <- "/media/nwknoblauch/Data/GTEx/GCTA_trans_ortho_estimates.RDS"
gcta_est <- readRDS(gcta_estf)
h_df <- filter(gcta_est,Source=="V(G)/Vp") %>% rename(h=Variance) %>% select(-Source)
h_elnet <- inner_join(h_df,lambda_min,by="fgeneid")
filter(h_elnet,h>1e-6) %>% ggplot(aes(x=h,y=lambda.min))+geom_point()+geom_smooth()