Trained PriLer models on European ancestry individuals

Prior learned elastic-net regression (PriLer) gene expression models trained on reference panels GTEx v6p and CMC release 1. Inside each folder, "genotype_info/" include tab-separated files divided per chromosome with info on variants used to train the models. "tissues/" include trained models divided per tissue. In the corresponding folders,<strong> resPrior_regEval_allchr.txt</strong> is a tab-separated file with summary statistics of trained model for each gene, sorted by chromosome and position. <strong>resPrior_regCoeffSnps_allchr.RData </strong>is an RData object including a Sparse Matrix per chromosome n. variants x n. of genes with trained regression coefficients. Variants position match genotype files and gene position match gene summary statistic order (divided per chromosome). <strong>ENSEMBL_gene_SNP_2e+05_chrxx_matrix.mtx </strong>are sparse 0/1 matrices per chromosome n. variants x n. of genes that represent gene-variant distance matrices. For each gene, 1 indicates a variant being located in the transcription starting site window +/- 200kb. To use the trained model to impute gene expression from on genotype dosages, follow Module 2 workflow in of https://gitlab.mpcdf.mpg.de/luciat/castom-igex.git