RegEvol: detection of directional selection in regulatory sequences through phenotypic predictions and phenotype-to-fitness functions

RegEvol finds 5.1% of Drosophila melanogaster regulatory regions show directional selection, with signals enriched near reproductive and immune genes.

• Why it matters: Selection in non-coding DNA is hard to detect because many methods use sequence conservation or substitution-rate patterns rather than molecular effects on regulation. RegEvol addresses this by connecting predicted regulatory impacts to explicit evolutionary models.
• What they did: It uses machine learning predictions of transcription factor binding to derive distributions of predicted mutational effects, then applies maximum-likelihood estimation to choose among drift, stabilizing selection, and directional selection along a lineage. It also aggregates likelihood differences across sets of regulatory elements when substitution counts are limited.
• The result: Across over 3 million Drosophila melanogaster regulatory regions, directional selection is inferred for 5.1% and is enriched near reproductive and immune genes. In humans, aggregating likelihoods for CTCF binding across tissues shows directional signals enriched in nervous and male reproductive systems.

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