Two Nature Biotechnology reports demonstrate deep‑learning and virtual screening approaches that accelerate antibacterial discovery by surfacing structurally novel compounds from high‑throughput datasets. Teams combined phenotypic screening data with machine learning models to prioritize candidates with antimicrobial activity and chemical novelty, shortening cycles of hit identification and de‑risking scaffold selection. Authors describe workflows that integrate experimental HTS results with model‑guided virtual screens, improving hit rates and enabling discovery of compounds outside classical antibiotic chemical space. The papers position AI‑augmented phenotypic strategies as a scalable complement to traditional medicinal chemistry in the race against antimicrobial resistance.