Direct discovery · Biology · Medicine · Computer science
Genomic language models generate viable bacteriophage genomes
AI-generated bacteriophage genomes that were synthesized and shown to infect bacteria in laboratory tests.
Summary
A bioRxiv preprint reports that genomic language models were used to generate complete bacteriophage genomes, with selected designs synthesized and tested for bacterial infection. The reported result connects sequence-generation models to a concrete biological output rather than only benchmark prediction.
AI role
Generated complete candidate viral genomes using genomic language models, after which selected designs were synthesized and experimentally assayed.
Narrative role
This backfills a 2025 biology gap for direct AI-generated biological systems and links the tracker from genomic foundation-model capability into experimentally tested genome design.
Caveat
The result is a preprint and should be treated as internally reported until peer review or independent replication; the work also has obvious biosecurity scope boundaries.