Announcing the BioML Challenge 2024: Bits to Binders! In this 5-week competition, teams will use AI methods to design protein binders that bind to a target antigen and activate an immune response from its associated CAR-T cell. Real-world scenario. With wet lab validation.

@leocastorina @DdelAlamo Ability to retrain it (imo). ProteinMPNN has training code but LigandMPNM does not (last I checked)

@gama_search Unless of course your model learns a perfect interconversion between any arbitrary SMILES representation. Then it would cause no variance. But from my experience this is unlikely given the number of small molecules that would be seen during training

And don't forget: When in doubt, canonicalize SMILES!

We need bioinformatics evals for LLMs. A lot of these problems have minimal & dated documentation yet are incredibly useful & should be simple Eg: Given a docked protein-ligand .cif file, split into two separate .pdbqt files Somehow none of the @OpenAI models can figure it out

You can learn more about the Bits to Binders protein design competition here:

RT @ClausWilke: Now published: We trained a chemical language model that works with both small molecules and peptides. Ideal for making pre…

@NikoMcCarty CFPS is the way

@leecronin How ?