Tracing human trait evolution through integrative genomics and temporal annotations

Abstract

Understanding the evolution of human traits is a fundamental yet challenging question. In a recent Cell Genomics article, Kun et al.¹ integrate large-scale genomic and phenotypic data, including deep-learning-derived imaging phenotypes, with temporal annotations to estimate the timing of evolutionary changes that led to differences in traits between modern humans and primates or hominin ancestors.

Understanding the evolution of human traits is a fundamental yet challenging question. In a recent Cell Genomics article, Kun et al. integrate large-scale genomic and phenotypic data, including deep-learning-derived imaging phenotypes, with temporal annotations to estimate the timing of evolutionary changes that led to differences in traits between modern humans and primates or hominin ancestors.

Main text

Natural selection has left distinct genomic signatures on the human genome. Advances in high-throughput sequencing technologies allow us to empirically investigate genomic differences across species and time points. However, discoveries of strong selective sweeps remain rare,² largely because (1) most human traits are complex, influenced by many variants with small effects,³ and (2) natural selection can adapt a population to an environmental change by subtly altering allele frequencies across many variants.⁴ These challenges make it difficult to trace the genetic evolution of complex traits.

One approach to identify genomic signatures of natural selection on complex traits is to aggregate trait-association signals within evolutionarily annotated regions. This requires (1) genome-wide association studies (GWASs), which map genetic variants associated with phenotypic variation of traits, and (2) genomic annotations, which provide information about functional roles of genomic regions or highlight sequence differences between species or populations. Statistical approaches to integrate and analyze these datasets include SNP-based heritability enrichment analysis⁵ and gene set enrichment analysis.⁶ An annotation is considered significant if SNPs within it, on average, explain a higher proportion of genetic variance than random SNPs in the genome or if there is an overrepresentation of genes associated with the trait (Figure 1). Overall, SNP-based heritability enrichment captures genome-wide signals but may be biased for annotations with small genomic lengths when using stratified linkage disequilibrium score regression (S-LDSC),⁵ while gene set enrichment focuses only on coding regions but is more robust to the annotation’s genomic length.

Figure 1.

Identification of genomic annotations associated with complex traits

This schematic illustrates the principle of statistical analysis, including SNP-based heritability and gene set enrichment approaches, for linking genomic annotations to trait variation. Each lollipop represents a SNP or a gene, with darker colors indicating a higher proportion of heritability explained by the SNP or a stronger trait-association signal of the gene. Colored circles indicate genomic annotations used in the analysis, which often have overlaps. In the study by Kun et al.,¹ the key annotations include human-gained enhancers and promoters (HGEPs), human accelerated regions (HARs), and ancient selective sweeps and Neanderthal-introgressed regions.

Kun et al. employed both SNP-based heritability enrichment (S-LDSC) and gene set enrichment (HARE, a pipeline they developed) approaches to investigate the timing of accelerated genomic changes. They focus on 70 complex human traits and diseases over the past 25 million years since divergence from rhesus macaques, chimpanzees, and Neanderthals and Denisovans. The GWAS data for selected traits were integrated with 11 evolutionary genomic annotations derived from comparative genomic and epigenetic studies across species and developmental stages, marking genomic regions that evolved during different periods of human evolutionary history. The authors observed, by and large, consistent but complementary results from S-LDSC and HARE.

The period of divergence between humans and rhesus macaques (∼25 million years ago [MYA]) was marked by human-gained enhancers and promoters (HGEPs). HGEPs were identified by comparing cis-regulatory activity of epigenetic elements between humans and rhesus macaques in limb and brain tissues across developmental stages. Significant heritability and gene enrichments were observed for skeletal traits, consistent with fossil evidence, highlighting adaptations related to bipedal locomotion and body structure. Additionally, signals were found for respiratory traits linked to lung function and white matter measurements in the brain’s left superior longitudinal fasciculus, which is associated with language processing.⁷

Human-chimpanzee divergence (∼5 MYA) was marked by human accelerated regions (HARs) and identified by comparing the human genome to mammalian and primate genomes. Significant enrichment was observed for body mass index (BMI), forced vital capacity (FVC), smoking status, neuroticism, and the visual cortex. Consistent with Xu et al.,⁶ Kun et al. identified a (non-significant) schizophrenia signal in HARs. A meta-analysis of traits by category revealed significant enrichment for psychiatric traits, aligning with prior evidence linking HARs to lung and brain development. This also expands our understanding of HARs’ potential roles in metabolic and psychiatric phenotypes during human development.

More recent human evolution (∼0.5 MYA) was marked by ancient selective sweeps and Neanderthal-introgressed regions (NIRs). S-LDSC analysis revealed heritability enrichment for autism in ancient selective sweeps, suggesting incomplete selection on autism-associated variants due to pleiotropic effects on neural development and cognition.⁸ In their HARE analysis, the authors found that immunological, dermatological, and respiratory traits were enriched in ancient selective sweeps, whereas reproductive and neurological traits showed enrichment in NIRs.

To account for the tissue specificity of epigenetic annotations, which were derived from fetal brain and limbs, Kun et al. performed additional analyses by restricting the evolutionary annotations to the cis-regulatory elements (CREs) active in the fetal brain. Consistent with the signals found in ancient selective sweeps, S-LDSC revealed enrichment for autism across all restricted evolutionary annotations. HARE analysis identified enrichment for respiratory function, balding, bone mineral density, and blood pressure CREs across all restricted evolutionary contexts. These results underscore the pleiotropic effects of active human enhancers and promoters across tissues.

In conclusion, a comprehensive study infers human evolutionary history for a range of complex traits by integrating GWAS data with temporal genomic annotations. This identified when accelerated evolution occurred for specific traits. While the results align with fossil evidence, they also revealed insights into the evolution of metabolic, immunological, respiratory, and psychiatric traits. This study, along with the authors’ previous work,⁹ pioneers the integration of deep-learning-derived imaging data with evolutionary annotations and represents the systematic linking of phenotypic evolution with distinct genomic changes over millions of years. These efforts lay a foundation for future research to uncover the evolutionary dynamics underlying human health and disease.

Looking ahead, additional GWAS datasets with increased sample sizes and diversity, alongside advances in functional genomics, will enable the creation of high-resolution annotations across diverse cell types and contexts. These developments will refine the temporal roadmap of human trait evolution. Furthermore, incorporating rare variants might improve our ability to infer the types of natural selection by contrasting the genetic variance explained by common and rare variants from different evolutionary periods.¹⁰

Acknowledgments

I acknowledge Professor Peter Visscher for helpful comments and support from the Australian National Health and Medical Research Council (1177268) and the Australian Research Council (DP220101947).

Declaration of interests

The author declares no competing interests.