PrediXcan 0% variant mapping issue

Author

Haky Im

Published

December 1, 2020

Many users had difficulties matching the genotype variant id to the prediction model variant id.

Here is one example added to the PrediXcan tutorial where the matching was failing because of the on the fly option not taking into account that in the GTEx v8 vcf file, chromosomes are names as chr# whereas in other vcf’s (more common in hg19?) chromosomes are indicated by their number or letter (no chr prefix).

A working example for the GTEx genotype data with GTEx v8 mashr models is shown below.

Code

export PRE=/gpfs/data/im-lab/nas40t2/Data/test-PrediXcan-GTEx
export DATA=$PRE/data
export MODEL=$PRE/models
export RESULTS=$PRE/results/
export METAXCAN=$PRE/repos/MetaXcan-master/software
export VCFSMALL=$PRE/data/gtex-small-common-test.vcf.gz

printf "Predict expression\n\n"

python3   $METAXCAN/Predict.py \
--model_db_path $PRE/models/gtex_v8_mashr/mashr_Whole_Blood.db \
--model_db_snp_key varID \
--vcf_genotypes  $VCFSMALL \
--vcf_mode genotyped \
--prediction_output $RESULTS/Whole_Blood__predict.txt  \
--prediction_summary_output $RESULTS/Whole_Blood__summary.txt \
--verbosity  \
--throw \
--on_the_fly_mapping METADATA "{}_{}_{}_{}_b38" 

** Thank you, Yanyu, for solving the mystery **

Reuse

© HakyImLab and Listed Authors - CC BY 4.0 for Text and figures - MIT for code