Abstract We are now at a pivotal point of medical and population genetics where available genetic and genomic datasets are powered to detect diverse signatures of natural selection on the human genome, and to investigate their downstream effects on the genetic architecture of human diseases and complex traits. Characterizing these signatures could enable us to improve our understanding of diseases, as well as their prevention, through improved polygenic risk scores across diverse ancestry groups, and diagnosis, through improved variant prioritization scores in clinical studies. However, methodological development and new analyses are still required to make sense of these new disparate datasets. In this proposal, we will develop models and apply methods aiming at investigate the downstream effects of natural selection on human diseases by leveraging novel large genetic and genomic datasets. First, we will characterize the genetic signatures of natural selection shaping the genetic architecture of human complex traits, by leveraging polygenic methods, genome-wide association studies (GWAS) summary statistics for a hundred of traits, and evolutionary simulations. Indeed, while many works have recently highlighted the action of negative selection on human diseases, we still need methods to analyze low-prevalence diseases and to investigate selection beyond the action of negative selection. Second, we will characterize the genetic signatures of recent selection leading to different gene regulation and allele effect sizes across diverse ancestry groups by leveraging single-cell RNA-seq and GWAS datasets from European and Asian ancestries. Developing methods to analyze and interpret the recently released non-European genomic and genetic datasets has the premise to understand recent human adaptation, and why allele effect sizes from GWAS differ across ancestry groups, which is fundamental to improve polygenic risk scores transportability. Third, we will characterize the genetic signatures of natural selection on genes at the exon and regulatory levels over millions of years of evolution by leveraging sequencing data from 240 mammals and recent enhancer-gene maps. Including the base pair resolution of constraint datasets to exon and regulatory scores will allow to improve our knowledge of gene evolution and function, and ultimately the interpretation of rare genetic variants in diagnostic studies. Our methods and datasets will be publicly available, deeply documented, and applicable to any heritable traits, maximizing their impact to the community.