De la Fuente lab uses ChatGPT, Codex to hunt new antibiotics

De la Fuente's lab is trying to solve a familiar bottleneck in drug discovery: finding a promising molecule takes years. The lab's answer is to treat biology as an information system. DNA and peptides are, in his words, like an alphabet, and deep-learning models can be trained to read patterns across genome and protein datasets far faster than manual screening. That approach can shrink the earliest search phase from years to hours.

The lab works across biology, chemistry, computer science, and engineering. ChatGPT and Codex help bridge those disciplines: biologists can write code, programmers can work on biological problems, and members can review unfamiliar topics or process large datasets in their native languages. De la Fuente also uses AI as a brainstorming partner, feeding it ideas from many lab members, though he cautions that accuracy must always be double-checked.

A promising candidate, however, is only a starting point. Before any molecule becomes a medicine, it must be confirmed to kill the target microbe, tested for safety and toxicity, assessed for how readily resistance develops, and manufactured reliably, then face regulatory review and clinical trials. The outcome hinges on that slow, grounded work joining the fast AI search—which is exactly why de la Fuente argues AI and laboratory biology must advance together.