Transcription start sites (TSSs) show a pronounced enrichment in GC-content, but the evolutionary processes generating this pattern remain unclear. We ask whether local mutation biases associated with transcription and replication and GC-biased gene conversion (gBGC) associated with recombination, can account for this variation without invoking selection.
Using population polymorphisms and de novo mutations from parent–offspring trios, we estimate mutation rates at single-nucleotide resolution around TSSs and compare observed nucleotide composition with mutation-based expectations. Forward simulations show that mutation alone does not reproduce the observed GC enrichment, whereas combining local mutation biases with gBGC captures its main features. An XGBoost analysis further identifies distinct groups of genes associated with different mutation rates, GC-content, and genomic features.
These results demonstrate the utility to  develop evolutionary models that can account for position-specific dynamics rather than gene averages. We aim to use an enriched framework to disentangle their contributions to nucleotide composition.