OpenAI Reveals Case Study Accelerating Antimicrobial Molecule Discovery with Codex and ChatGPT

OpenAI released a case study in which César de la Fuente's lab utilized Codex and ChatGPT to discover new antimicrobial molecules. With deaths from antimicrobial resistance (AMR) projected to nearly double by 2050, traditional antibiotic development had faced diminishing returns due to the absence of new classes.

AI-Based Discovery Methodology

The research team treated biology as an information system, likening DNA and protein sequences to alphabets, and used deep learning models to recognize patterns. This allowed them to search for antimicrobial candidates from genome data of extant and extinct organisms, reducing the initial discovery time from years to hours.

Tool Usage and Collaboration

Codex and ChatGPT were used for hypothesis brainstorming, code writing, dataset processing, and connecting interdisciplinary ideas. In particular, they lowered barriers between fields, enabling biologists to build programs and programmers to solve biological problems. ChatGPT played a role in mitigating language barriers and accelerating workflows.

Validation and Limitations

Even if AI identifies promising candidates, they are not confirmed as drugs. Ground-truth experiments are essential, including confirming the killing of target microbes, determining effective concentrations, and testing effects on human cells. Subsequently, complex processes such as chemical optimization, toxicity evaluation, and clinical trials must be undergone. AI is establishing itself as a tool to help break through academic boundaries, but accuracy must always be verified.