The abstract is read at the publisher; the summary is JClub's.
Randomly assigned terminal trait labels on presumed-null single-copy genes are rejected by PAML’s two-ratio branch model in 65.9–88.8% of genes across three fungal/yeast/lactobacilli panels, far above 5%.
- Why it matters: Branch-partitioned selection tests are often interpreted using nominal nulls, but it is unclear whether they truly behave like random labeling when the trait signal is absent.
- What they did: In three ~50-species panels, terminal labels were randomly placed on presumed-null genes and tested with PAML’s two-ratio branch model; refitting nulls per gene and simulating at the alignment depth still left 65.0% and 83.8% rejection for 80 genes.
- The result: Branch-site tests rejected none after refitting, and correct specification at the data depth gave 5.4–6.8% rejection, implying trait labels should be benchmarked against random labels of the same size rather than chi-squared.
The findingWhy it mattersWhat they didThe result
AI summary of the abstract · may contain errors · not medical advice · read the paper.