Improved detection of evolutionary selection highlights potential bias from different sequencing strategies in complex genomic-regions

Balancing selection occurs when multiple alleles are kept at elevated frequencies in equilibrium due to opposing evolutionary pressures. A new statistical method was developed to test for selection using efficient Bayesian techniques. Selection signals in three different data sets, generated with variable sequencing technologies, were compared: clinical trios, HLA NGS typed samples, and whole-genome long-read samples. Genome-wide, selection was observed across multiple gene families whose biological functions favor diversification, revealing established targets as well as 45 novel genes under selection. Using high-resolution HLA typing and long-read sequencing data, for the characterization of the MHC, revealed strong selection in expected peptide-binding domains as well as previously understudied intronic and intergenic regions of the MHC. Surprisingly, SIRPA, demonstrated dramatic selection signal, second only to the MHC in most settings. In conclusion, employing novel statistical approaches and improved sequencing technologies is critical to properly analyze complex genomic regions.