Our work addresses a fundamental question in genetics and microbiology, the accurate inference of traits (phenotypes) from genome sequence (genotypes). To enable accurate predictions of bacterial metabolic capabilities, gene regulatory networks, and stress responses from genome sequence alone, we are developing high-throughput genetic tools that can be applied inexpensively to a large number of bacteria and generating large gene-phenotype datasets for gene function inference. In this talk, I will discuss our progress in developing these genetic tools, and how we are integrating these data with comparative genomic methods to identify new genes involved in carbon catabolism, amino acid biosynthesis, and host colonization.