OpenAI: Codex and ChatGPT help hunt antimicrobial candidates

OpenAI published how César de la Fuente’s lab uses Codex and ChatGPT — plus its own deep-learning models — to search living and extinct genomes for antimicrobial candidates, which are early discovery hits, not approved drugs.

OpenAI put its coding and chat tools inside a real hunt for new infection-fighting molecules. A faster genome search is not a medicine that has survived the lab and the clinic.

OpenAI published the Applied AI story “How a researcher uses Codex and ChatGPT to search for new antimicrobial molecules.” That company page is the primary.

OpenAI profiles César de la Fuente’s lab. The lab, it says, searches the genomes of living and extinct organisms for molecules that could fight drug-resistant infections. Antimicrobial means a drug that kills or stops microbes. De la Fuente is quoted calling antimicrobial resistance “one of the greatest existential threats to humanity in my opinion.” He also says there has not been a new class of antibiotics for 50 years. Those are his comments as published by OpenAI.

OpenAI’s page cites about five million deaths in 2021 tied to bacterial antimicrobial resistance. It says that yearly toll is projected to roughly double by 2050, linking The Lancet. Those figures are on OpenAI’s post — not desk-verified epidemiology.

OpenAI says finding molecules that might become antimicrobials can take years. The lab’s own deep-learning models — software trained to spot patterns in biological sequences — can cut that first search from years to hours, the page says. Treat the speed-up as company and lab framing, not an independent bench.

The lab also uses ChatGPT and Codex to brainstorm ideas, write and refine code, process datasets, analyze results, and bridge biology, chemistry, and computer science. De la Fuente, on the page: ChatGPT is a collaborative workspace that takes input from people who think differently about the problems. He adds, “you have to always double-check for accuracy.”

A candidate is not an approved drug. OpenAI’s page walks the rest of the path: confirm the molecule kills the target microbe; dose; effects on human cells; chemical tweaks; toxicity; how easily microbes resist it; how it moves through the body; a reliable way to make it; then regulatory review and clinical trials. De la Fuente: “Ground-truth experiments are essential to validate AI predictions.” This filing does not name molecules, trial results, FDA approvals, or product SKUs beyond Codex and ChatGPT as described.