Check out our new preprint!
How can we get more out of existing protein/genome language models without retraining them?

Awesome paper. Simple post-hoc trick (averaging over homologs) with elegant evolution theory dramatically improves zero shot coding variant effect prediction in pLMs & actually delivers better results from the larger models (inverting the trend with raw likelihoods). 1/

@cwjpugh.bsky.social at #VariantEffect25

Proud supervisor moment! Next up at #VariantEffect25: Charles Pugh @cwjpugh.bsky.social, PhD student in Mafalda’s and my group, presents: From Likelihood to Fitness: Improving Variant Effect Prediction in Protein and Genome Language Models #AI #ProteinML

A lovely summary of what we have learnt about scaling from ProteinGym -- Have we reached a performance plateau? More on this later in the week from @cwjpugh.bsky.social