Nick Mancuso 2018-02-28
RHOGE is an R package that estimates the genome-wide genetic correlation between two complex traits (diseases) as a function of predicted gene expression effect on trait (\rho_{ge}). Given output from two transcriptome-wide association studies, RHOGE estimates the mediating effect of predicted gene expression and estimates the correlation of effect sizes across traits (diseases). This approach is extended to a bi-directional regression that provides putative causal directions between traits with non-zero \rho_{ge}.
This approach is described in:
Integrating Gene Expression with Summary Association Statistics to Identify Genes Associated with 30 Complex Traits Nicholas Mancuso, Huwenbo Shi, Pagé Goddard, Gleb Kichaev, Alexander Gusev, Bogdan Pasaniuc. American Journal of Human Genetics. 2017.
Bioconductor + devtools is the most straightforward way to install RHOGE. To do this open an R terminal and enter
source("http://bioconductor.org/biocLite.R")
biocLite("devtools") # only if devtools not yet installed
biocLite("bogdanlab/RHOGE")The following example computes \rho_{ge} between BMI and triglycerides, as well as putative causal directions.
library(RHOGE)
# example BMI TWAS results from FUSION
data(bmi)
head(bmi)## # A tibble: 6 x 22
## FILE ID CHR P0 P1 HSQ BEST.GWAS.ID BEST.GWAS.Z EQTL.ID
## <chr> <chr> <int> <dbl> <dbl> <dbl> <chr> <dbl> <chr>
## 1 /u/ho… LOC6… 1 7.63e5 7.95e5 0.0979 rs17160824 1.96 rs3094…
## 2 /u/ho… AGRN 1 9.56e5 9.91e5 0.0655 rs17160824 1.96 rs9442…
## 3 /u/ho… C1or… 1 1.02e6 1.05e6 0.0917 rs17160824 1.96 rs3766…
## 4 /u/ho… SCNN… 1 1.22e6 1.23e6 0.170 rs9660180 4.89 rs1126…
## 5 /u/ho… MXRA8 1 1.29e6 1.30e6 0.0503 rs9660180 4.89 rs1739…
## 6 /u/ho… LOC1… 1 1.33e6 1.34e6 0.0648 rs9660180 4.89 rs2649…
## # ... with 13 more variables: EQTL.R2 <dbl>, EQTL.Z <dbl>,
## # EQTL.GWAS.Z <dbl>, NSNP <int>, NWGT <int>, MODEL <chr>,
## # MODELCV.R2 <dbl>, MODELCV.PV <dbl>, TWAS.Z <dbl>, TWAS.P <dbl>,
## # PERM.PV <dbl>, PERM.N <int>, PERM.ANL_PV <dbl># example triglyceride TWAS results from FUSION
data(tg)
head(tg)## # A tibble: 6 x 22
## FILE ID CHR P0 P1 HSQ BEST.GWAS.ID BEST.GWAS.Z EQTL.ID
## <chr> <chr> <int> <dbl> <dbl> <dbl> <chr> <dbl> <chr>
## 1 /u/ho… LOC6… 1 7.63e5 7.95e5 0.0979 rs4314833 2.32 rs3094…
## 2 /u/ho… AGRN 1 9.56e5 9.91e5 0.0655 rs4314833 2.31 rs9442…
## 3 /u/ho… C1or… 1 1.02e6 1.05e6 0.0917 rs11552172 -2.36 rs3766…
## 4 /u/ho… SCNN… 1 1.22e6 1.23e6 0.170 rs6604981 2.44 rs1126…
## 5 /u/ho… MXRA8 1 1.29e6 1.30e6 0.0503 rs13303010 2.60 rs1739…
## 6 /u/ho… LOC1… 1 1.33e6 1.34e6 0.0648 rs13303010 2.89 rs2649…
## # ... with 13 more variables: EQTL.R2 <dbl>, EQTL.Z <dbl>,
## # EQTL.GWAS.Z <dbl>, NSNP <int>, NWGT <int>, MODEL <chr>,
## # MODELCV.R2 <dbl>, MODELCV.PV <dbl>, TWAS.Z <dbl>, TWAS.P <dbl>,
## # PERM.PV <dbl>, PERM.N <int>, PERM.ANL_PV <dbl># Estimate rho_ge genome-wide for BMI and Triglyerides and approximate sample sizes
ge_cor_res <- rhoge.gw(bmi, tg, 14000, 91000)
head(ge_cor_res)## RHOGE SE TSTAT DF P
## 1 0.1949802 0.06210183 3.139686 461 0.001799913# Perform bi-directional regression to estimate putative causal directions
bidir_res <- rhoge.bd(bmi, tg, 14000, 91000, p1 = 0.05 / nrow(bmi), p2 = 0.05 / nrow(tg))
head(bidir_res)## ESTIMATE SE TSTAT DF P TEST
## 1 0.560464 0.1043080 5.3731637 35.00000 5.180204e-06 Trait1 -> Trait2
## 2 -0.117688 0.2305352 -0.5104989 35.00000 6.129068e-01 Trait2 -> Trait1
## 3 NA NA 2.6800731 50.14695 9.932761e-03 DIFFCurrently, only FUSION style output is supported.
RHOGE comes installed with estimates of approximately independent LD blocks for European, Asian, and African ancestries. Performance should improve if you have in-sample estimates of LD blocks. The only requirement is that regions are stored as a data.frame-like object with 3 columns ('CHR', 'START', 'STOP'). For example,
library(RHOGE)
data("grch37.eur.loci")
head(grch37.eur.loci)## # A tibble: 6 x 3
## CHR START STOP
## <int> <int> <int>
## 1 1 10583 1892607
## 2 1 1892607 3582736
## 3 1 3582736 4380811
## 4 1 4380811 5913893
## 5 1 5913893 7247335
## 6 1 7247335 9365199