Penn Lab Breakthrough Uses AI to Unlock New Antibiotic

Researcher César de la Fuente at the University of Pennsylvania has used AI to identify new antimicrobial peptide candidates capable of fighting drug-resistant infections.

Key Takeaways

Antibiotic resistance kills more than a million people a year worldwide
César de la Fuente’s lab at the University of Pennsylvania mines living and extinct genomes for antimicrobial peptide candidates
Researchers use Codex and ChatGPT to write and iterate on search scripts for antimicrobial discovery
OpenAI’s post does not name specific molecules discovered or disclose clinical timelines

His lab mines both living and extinct genomes for peptide sequences with antibiotic potential, according to a post OpenAI published Thursday. The approach treats genetic databases as an untapped search space rather than a static archive, using AI models to flag molecular patterns human researchers would take years to sift through manually.

Why Penn And Others Turned Drug Resistance Into A Search Problem

Antibiotic resistance kills more than a million people a year worldwide, and the pipeline of new drugs has thinned as pharmaceutical companies pulled back from antibiotic research over the past two decades because returns are lower than for chronic-disease drugs.

De la Fuente’s lab has published on mining extinct organisms’ DNA, including data drawn from fossils and ancient samples, on the theory that molecules from species long gone may show antibacterial properties unlike anything in circulation today.

Also Read: DeepSeek’s Latest Breakthrough Crushes AI Pricing Power

From Manual Screening To Machine-Scale Search

Using Codex and ChatGPT, researchers can write and iterate on search scripts quickly, while ChatGPT handles analysis and hypothesis generation.

Antimicrobial peptides are short chains of amino acids that many organisms produce naturally to kill bacteria, and resistant strains have historically struggled to develop defenses against them, unlike conventional antibiotics.

This sits inside a broader field called digital biology, where AI models search genomic and proteomic data the way a search engine indexes web pages.

Traditional antimicrobial discovery involved testing compounds one at a time in wet labs, a process that could take a graduate student years to screen a few thousand candidates. This mirrors a wave of AI-driven biology projects that surfaced across the industry through 2026, as labs increasingly positioned coding assistants as general-purpose research tools rather than programming aids alone.

What OpenAI Says Comes Next

OpenAI’s post does not name specific molecules discovered or disclose clinical timelines, leaving open how close any candidate is to trials.

The company positioned the case study as evidence that agentic coding tools can accelerate scientific search broadly, a claim it has made before pointing to fields including materials science.