Codex and ChatGPT Aid Search for New Antimicrobials

Researcher César de la Fuente is leveraging OpenAI's Codex and ChatGPT to search through living and extinct genomes for antimicrobial molecules that could combat drug-resistant infections. His lab's work represents a breakthrough application of large language models in computational biology, where AI accelerates the discovery of peptide sequences with therapeutic potential. As antibiotic resistance threatens to become one of the century's most urgent health crises, this approach offers a scalable pathway to identify novel compounds buried in genomic data.

The Drug Resistance Challenge

Antimicrobial resistance has reached alarming levels globally. The World Health Organization estimates that drug-resistant infections could cause 10 million deaths annually by 2050 if no action is taken. Traditional drug discovery methods are time-consuming and expensive, often taking over a decade to bring a single compound from laboratory to clinic. De la Fuente's lab addresses this bottleneck by using AI to rapidly screen genomic sequences for antimicrobial candidates, filtering millions of potential molecules in a fraction of the time required by conventional methods.

How Codex and ChatGPT Analyze Genomes

The research team employs Codex and ChatGPT to parse complex genomic datasets from both living organisms and extinct species preserved in fossil records. Codex, originally designed for code generation, excels at pattern recognition in structured data, making it ideal for identifying peptide sequences with antimicrobial properties. ChatGPT assists in hypothesis generation and literature synthesis, helping researchers contextualize findings within existing scientific knowledge.

The workflow begins with data extraction from public genomic databases. The AI models then scan for specific sequence patterns known to correlate with antimicrobial activity, such as cationic peptides that disrupt bacterial membranes. By analyzing extinct genomes, the lab can uncover ancient defense mechanisms that may have been lost through evolution but remain effective against modern pathogens.

Release Date and Availability

This research application was officially launched on September 10, 2026, when OpenAI published the case study on their blog. The methodology is currently being used in de la Fuente's lab at the University of Pennsylvania, where it has already identified several promising antimicrobial candidates now undergoing laboratory validation. While the specific models (Codex and ChatGPT) are generally available through OpenAI's API, the custom workflows and training data used in this research remain proprietary to the lab.

Early Results and Validation

Initial findings from the lab have been encouraging. The AI-assisted approach has identified peptide sequences from Neanderthal and mammoth genomes that show activity against drug-resistant Staphylococcus aureus and Pseudomonas aeruginosa strains. The team reports that their computational screening reduces the candidate pool from millions to hundreds, which are then synthesized and tested in vitro. This filtering process increases the hit rate for active compounds by an estimated 40% compared to random screening.

• Scans living and extinct genomic databases for antimicrobial peptide patterns
• Reduces discovery timeline from years to months through rapid computational filtering
• Identifies candidates effective against multiple drug-resistant bacterial strains
• Combines Codex pattern recognition with ChatGPT hypothesis generation

What This Means

De la Fuente's work demonstrates how general-purpose AI models like ChatGPT and Codex can be adapted for highly specialized scientific domains without extensive retraining. The success of this approach suggests that AI-driven drug discovery will become increasingly accessible to research labs worldwide, potentially democratizing the search for new therapeutics. As computational power and genomic databases continue to expand, the intersection of AI and biology may finally provide the tools needed to outpace evolving microbial threats. The next phase will focus on optimizing identified peptides for clinical use and expanding the search to viral and fungal pathogens.