Predicting metal-protein interactions using cofolding methods: Status quo
Duerr, S. L.; Rothlisberger, U.
Show abstract
Metals play important roles for enzyme function and many therapeutically relevant proteins. Despite the fact that the first drugs developed via computer aided drug design were metalloprotein inhibitors, many computational pipelines for drug discovery still discard metalloproteins due to the difficulties of modelling them computationally. New "cofolding" methods such as AlphaFold3 (AF3) (Abramson et al., 2024) and RoseTTAfold-AllAtom (RFAA) (Krishna et al., 2024) promise to improve this issue by being able to dock small molecules in presence of multiple complex cofactors including metals or covalent modifications. Here, we analyze the current status for metal ion prediction using these methods. We find that currently only AF3 provides realistic predictions for metal ions, RFAA in contrast does perform worse than more specialized models such as AllMetal3D in predicting the location of metal ions accurately. We find that AF3 predictions are consistent with expected physico-chemical trends/intuition whereas RFAA often also predicts unrealistic metal ion locations.
Matching journals
The top 4 journals account for 50% of the predicted probability mass.
Similar papers in this journal
"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.