Incorporating tandem repeats improves gene expression prediction models

Adding tandem repeat (TR) copy-number features to cis-window gene expression models improves prediction, yielding significant gains for 78 genes and a mean adjusted R-squared increase of 0.0037.

• Why it matters: TRs are common and polymorphic, yet most gene expression prediction models ignore them, leaving a gap when causal TRs are poorly captured by SNPs.
• What they did: The authors built SNP+TR expression prediction models using Geuvadis lymphoblastoid cell line data (n=359 European individuals), comparing against SNP-only models across 2,553 successfully trained genes.
• The result: SNP+TR models improved performance for 78 genes, showed mean adjusted R-squared gain 0.0037 across all genes, replicated in GTEx, and nonlinear TR testing identified 2,255 nonlinear significant genes.

AI summary of the abstract · may contain errors · not medical advice · read the paper.