Artificial intelligence is enhancing and accelerating our ability to extract biological insight from sequence data. In this seminar, two complementary applications of artificial intelligence that operate at different biological scales are presented. At the genomic level, machine learning applied to Salmonella enterica genomes reveals that gene presence–absence patterns can accurately distinguish virulent from attenuated Salmonella serovars and identify key genetic drivers of pathogenicity, with direct implications for public health and risk assessment. At the molecular level, a protein-coevolution-guided AlphaFold3 pipeline enables the prediction of protein–protein interactions by prioritizing biologically meaningful candidates and reconstructing their structures in silico, with particular relevance for complexes involved in rare diseases, such as the BBSome, with the aim of identifying novel protein complexes and new interactors of known complexes. Together, these approaches illustrate how artificial intelligence uncovers hidden relationships: from genes to phenotypes and from evolutionary signals to molecular interactions.