paperKB
coga / coga-kb
Help
Sign in

Chunk #14 — Results — Inferring gene expression from L1000 landmarks

Source
A Next Generation Connectivity Map: L1000 Platform and the First 1,000,000 Profiles.
Embedded
yes

Text

Using 8,555 RNA-seq samples (Dataset DSGTEx-rnaseq) as an independent test set, we used landmark transcript measurements to infer the remainder of the transcriptome. As a test of inference accuracy, we analyzed gene-level recall (Rgene) for each of the inferred genes and assessed performance by comparing the result to a null distribution of correlations between all inferred transcripts and all measured transcripts. This analysis showed that inference was accurate (defined as Rgene > 0.95) for 9,196 of the 11,350 inferred genes (81%). When combined with the 978 measured landmarks, the L1000 platform thus measures or infers with high fidelity 83% of transcripts, but yields poor inference for 17% (Figure 1E, lower panel right and Table S3). Inferences for these 17% were therefore not used in any of the analyses that follow.