A longstanding promise of biology is that with a deep enough understanding of the molecular rules of life, it should be possible to predict phenotypes from genotype and environment. This would enable rational genome engineering of crops and microbes for desirable traits and facilitate precision medicine through interventions based on an individual’s genetic variation and lifestyle. To make progress towards these goals, technologies are needed to comprehensively identify causal variants, their effects on genes and their interactions, directly in natural populations. Recent advances in high-throughput genome editing make this possible in the model eukaryote S. cerevisiae. This proposal will employ a high-efficiency, multiplexed genome editing system (MAGESTIC) where each cell in a pool receives a distinct edit for a natural variant and a corresponding barcode, which is integrated into the genome after editing. The barcode allows for building arrayed collections of validated strains by recombinase-directed indexing (REDI) and for tracking variant abundance in pooled growth assays by next-generation sequencing (NGS). To identify the basis for trait variation across the S. cerevisiae lineage, five distinct strain backgrounds derived from genetically and ecologically diverse wild isolates will be employed. Libraries will be designed for each strain to introduce >90% of the 85,000 variants observed in the other four strains, enabling studying the impact of the major and minor alleles and the background dependence of their effects. The variant pools will be assayed across a panel of conditions relevant to pathogenic fungi, wild isolates, and human disease, as many variants are expected to exert their effects only in certain environments. To understand how multiple variants modify multigenic traits, a sequential editing and barcoding technology (MARVEL) will be used to generate pairwise and higher-order combinations for up to hundreds of causal variants per trait. This approach will characterize the extent of non-additive effects between variants (i.e. genetic interactions) and genetic background dependencies. This effort will constitute the most comprehensive investigation of genotype-environment-phenotype relationships across a species, and of wild isolates, to date. Ultimately, the data and insights generated by our study will facilitate predictive models linking variants to pathways and phenotypes beyond the S. cerevisiae system, and in particular to other organisms where precision CRISPR editing at this scale is not yet feasible.