Chunk #50 — Methods — Estimation of the genetic component of gene expression levels (GReX) — Performance of transcriptome prediction in independent cohorts
We tested the feasibility of predicting the transcriptome (i.e., estimating the genetic component of each gene expression trait, GReX^ in an independent test transcriptome dataset) using the elastic net effect sizes trained in the DGN whole blood data (n=922). For the test sets, we used independent RNA-Seq datasets from 421 LCL cell lines from the 1000 Genomes project generated by the GEUVADIS consortium15 and the nine tissues from the GTEx pilot project14 (see Supplemental Figure 3). To assess performance, we used the square of the Pearson correlation, R2, between predicted and observed expression levels.
