Most of the peptide conversation focuses on compounds that already exist — BPC-157, semaglutide, CJC-1295.
But behind the scenes, the way peptides are discovered is changing completely.
What happened:
Multiple publications in March 2026 confirm that AI has shifted from predicting which existing peptides might have antimicrobial activity to actively designing new peptide candidates from scratch.
A pipeline called ProteoGPT, published in Nature Microbiology, enables high-throughput screening across hundreds of millions of peptide sequences.
A separate review in JACS Au mapped how machine learning and deep learning approaches are producing experimentally validated antimicrobial peptides effective against multidrug-resistant pathogens.
A third study in Scientific Reports used AI topic models to identify structural motifs that determine antimicrobial function — essentially finding the building blocks that make a peptide effective.
Why this matters for biohackers:
The compounds being discussed today represent a fraction of what’s coming. AI-designed peptides are already entering early validation phases.
The rate at which new candidate compounds move from concept to testing is accelerating dramatically.
If you only pay attention to the compounds that are trending right now, you’re watching the rearview mirror.
Understanding how peptides are discovered — not just what they do — is part of building genuine research literacy.
We discuss emerging research like this regularly inside the community.
— The Biohacker Network