Machine-learning dissection of Human Accelerated Regions in primate neurodevelopment

Abstract

Using machine learning (ML), we interrogated the function of all variants in 714 Human Accelerated Regions (HARs), some of the fastest evolving regions of the human genome. We predicted that 38% of HARs have variants with opposing effects on neurodevelopmental enhancer activity, consistent with compensatory evolution, and we confirmed this with massively parallel reporter assays (MPRAs) in human and chimpanzee neural progenitor cells. Our ML analysis also revealed 159 differentially active HARs that can be predicted from species-specific transcription factor footprints. Despite these striking cis effects, activity of a given HAR sequence was nearly identical in human and chimpanzee cells. These findings suggest that HARs did not evolve to compensate for changes in the trans environment but instead altered their ability to bind factors present in both species. Thus, ML revealed an unexpected reason why HARs evolved so rapidly and prioritized variants with functional effects on human neurodevelopment.