klkeys / geuvadis-predixcan

Pipeline to train and analyze PrediXcan prediction weights using GEUVADIS EUR and YRI populations

geuvadis-predixcan

Pipeline to train and analyze PrediXcan prediction weights using data from the Genetic European Variation in Health and Disease (GEUVADIS) project.

Genotypes and normalized gene expression measures for 373 EUR samples and 89 YRI samples from GEUVADIS were downloaded from the EBI website.

Per protocol from Gamazon et al. (2015), new prediction weights for each population were computed using elastic net regularized linear regression with the glmnet package in R.

Steps

1. Download and parse GEUVADIS data (docs)
2. Train and test prediction models using continental populations (docs)
3. Train and test prediction models using the five constituent GEUVADIS populations (docs)