As AI (that is to say, LLM-based technology, until the current trend drastically changes) continues to mature and develop, it’s interesting to think about its impact on different domains beyond software engineering where it is arguably having the most visible impact. I have spent my professional career as a biologist and undeniably there have been enormous AI growth in my field. Given the pace of LLM development, this post will likely remain perpetually out of date.

I think it’s useful to frame these AI developments as enhancing the capabilities of biologists rather than replacing them outright, as very well articulated in this article from May 2026.

Answering biomedical questions

Since the most popular implementation of AI is chatbots, it’s hardly surprising that there have been many attempts to implement AI biology expert systems. Med-PaLM and its follow-up Med-PaLM 2 can answer medical questions drawn from the US Medical Licensing Exam (USMLE). Heidi can answer clinical questions.

Biological foundation models

Analogous to foundation large language models, there are biological foundation models trained on biological sequences. Evo was trained on bacterial genomes and Evo 2 was trained on all domains of life. Models for more specific tasks involving DNA or protein sequences are then built upon these foundation models.

Modeling biological processes using AI

Deep learning have been applied to H&E slides to predict mutations (DeepPATH), molecular phenotypes (PACpAInt), transcriptomic levels (HE2RNA) and patient outcomes (MesoNet).

For histology there is H0-mini, a distillation of U-Optimus and M-Optimus foundational models.

Modeling cell surface receptors, such as Trex for T cells and Ibex for B cells.

For designing mRNA sequence there is mRNABERT.

For protein-ligand docking there are DiffDock, IsoDDE, Boltz-2, ESMFold.