Direct discovery · Biology · Medicine
ProGen generates functional protein sequences across families
Artificial protein sequences generated by a language model and tested for function.
Summary
The Nature Biotechnology paper describes ProGen, a protein language model that generates sequences across diverse protein families. The authors experimentally test artificial lysozymes and report functional proteins with sequence identities substantially below close natural homologues.
AI role
Generated protein sequences across families using language-model training on biological sequence data.
Narrative role
ProGen is an early generative-biology milestone because it shows language models proposing functional sequences beyond close natural templates.
Caveat
Functional testing was limited to selected examples and does not imply broad protein-design reliability.