How much do routine RNA-seq processing decisions affect downstream transcriptomics analyses?
“Quite a bit, as it turns out,” said Nolan Cole, a University of Washington biostatistics doctoral student and co-author of a paper recently published in Nature Genetics.
Across 48 GTEx tissues, researchers found that changing the expression quantification method or transcript annotation resulted in substantial differences in eQTL discovery, gene expression prediction, and ultimately which genes were prioritized through GWAS colocalization and TWAS.
Cole highlighted some of the team’s striking results:
- 39-43% discordance in eGene identification across quantification methods,
- 32-39% discordance across transcript annotations, and
- even greater discrepancies in downstream gene prioritization, sometimes identifying different candidate genes at the same GWAS locus.
“The broader takeaway is that we need better expression measurements and references, particularly through long-read sequencing and the finer resolution of data. Reproducibility in science requires paying attention to both how we generate data, not just the downstream statistical methods,” said Cole.
The study was led by University of Texas MD Anderson Cancer Center researchers Taylor Head and Arjun Bhattacharya.
