A pair of Nature Biotechnology papers demonstrate that deep‑learning models trained on high‑throughput phenotypic and screening datasets can identify novel antibacterial scaffolds. One study combined phenotypic screening with AI‑driven virtual screens to nominate compounds active against multidrug‑resistant pathogens; the other trained models on experimental HTS data to find structurally novel antimicrobials. Both reports highlight mechanisms to scale antibacterial discovery and prioritize chemically distinct hits for follow‑up, addressing a critical void as resistance outpaces conventional discovery pipelines.