README.md

In GMELab/GraBLD: Gradient Boosted and LD adjusted

GraBLD

Gradient boosted and LD adjusted (GraBLD) is an R based software package that applied to polygenic traits prediction using gene scores.

Quick Start###

The current release is: version 1.0. See "release" tab.

To install our R package “GraBLD”, you can either run in R directly:

install.packages("devtools") # if you have not installed already
devtools::install_github("GMELab/GraBLD")

OR

After downloading GraBLD_0.1.0.tar.gz at "release" tab, run:

R CMD INSTALL GraBLD_0.1.0.tar.gz

To load the library in R, simply run in R:

library(“GraBLD”)

Tips for using GraBLD on large datasets###

The package contains two main functions, one is to calcualte the LD adjustments and the other to calculate the gradient boosted polygenic score weights. The two are combined in the end to give the actual weights of the polygenic score.

LD adjustments##

For large datasets, it is recommended to run from the command line with

for((i = 1; i <= chr; i++))
do
Rscript PerformLDadj.R size data_name ${i} &
done

where the R script "PerformLDadj.R" might look something like this, while additional options can be added to the argument list:

#!/bin/sh
rm(list = ls())
library('GraBLD')
args = (commandArgs(TRUE))
size = eval(parse(text=args[1]))
source_data = args[2]
chr = eval(parse(text=args[3]))
geno_data = load_geno(source_data = source_data, PLINK = TRUE, chr = chr)
geno_norm = full_normal_geno(geno_data)
LD_OUT <- LDadj(geno_raw = geno_norm, chr = chr, size = size, write = TRUE)

Gradient boosted weights##

For large datasets, it is recommended to run from the command line with

geno_data="chr1.raw" # or the actual file name of the genotype data 
trait_name="BMI" # name of the trait
annotations_file="annotation.txt" # or the actual file name of the annotation data
validation=5 # or value of your choosing
interaction_depth=5 # or value of your choosing
shrinkage_parameter=0.01 # or value of your choosing
bag_fraction=0.5 # or value of your choosing
maximum_tree=2000 # or value of your choosing
for (( i = 1; i <= $validation; i++))
do
Rscript calculate_gbm.R $geno_data
   $trait_name $annotations_file ${i}
   $validation $interaction_depth
   $shrinkage_parameter $bag_fraction
   $maximum_tree &
done

where the R script calculate_gbm.R might look something like this, while additional options can be added to the argument list:

#!/bin/sh
rm(list = ls())
library('GraBLD')
args = (commandArgs(TRUE))
geno_data = args[1]
trait_name = args[2]
annotations_file = args[3]
steps = eval(parse(text=args[4]))
validation = eval(parse(text=args[5]))
p1 = eval(parse(text=args[6]))
p2 = eval(parse(text=args[7]))
p3 = eval(parse(text=args[8]))
p4 = eval(parse(text=args[9]))
betas = load_beta(trait_name)
annotation = load_database(annotations_file, pos = 2:3) # taking the 2nd and 3rd columns of "annotations_file"
geno <- load_geno(geno_data)
GraB(betas = betas, annotations = annotation,
  trait_name = trait_name, steps = steps, validation = validation,
  interval = 200, sig = 1e-05, interact_depth = p1, shrink = p2,
  bag_frac = p3, max_tree = p4, WRITE = TRUE)

Combining the two##

The two steps above produce separate files, one for the LD adjustments and the other the boosted weights, and they can be combined by the function "GraBLD.score".

LD_val <- read.table("OUTPUTS_FROM_STEP1.txt") ## if each chromosome is computed separately, you will need to append them into a single file 
gbm_val <- read.table("OUTPUTS_FROM_STEP2.txt")
gs <- GraBLD.score(geno_raw = YOUR_GENO_DATA, LDadjVal = LD_val, gbmVal = gbm_val, Pheno = NULL)

Notice that if "Pheno" is supplied, both the polygenic gene score as well as the prediction R-squared (adjusted) are returned, otherwise only the polygenic gene score is returned.

GMELab/GraBLD documentation built on May 4, 2019, 3:20 p.m.