A research lab led by bioengineer César de la Fuente is using AI to search genome and protein databases for molecules that might fight drug-resistant microbes. Its deep-learning models look for useful patterns in biological sequences, reducing the initial hunt for candidates from years to hours, according to an OpenAI case study.
The lab also uses ChatGPT and Codex to brainstorm hypotheses, write and refine code, process datasets and connect ideas across disciplines. The broader strategy treats DNA and protein sequences as information that can be searched across living and extinct organisms, rather than limiting discovery to familiar chemical families.
That speed applies only to the earliest discovery stage. A predicted molecule must still be tested against the target microbe and human cells, optimized for safety and stability, assessed for toxicity and resistance, manufactured reliably, and eventually pass regulatory review and clinical trials. The researchers stress that laboratory experiments provide the ground truth for model predictions.
The work illustrates where general AI assistants can complement specialized scientific models: they can help researchers explore and analyze a much larger search space, but they do not replace the experimental evidence required to turn a candidate into a treatment.