While recent co-folding models such as AlphaFold-3 achieve accurate structure prediction, they fail to generalize to underexplored binding interfaces - systematically misplacing ligands, particularly

The gap

Consensus across the literature

Supporting evidence — 3 representative gaps

• Structure and dynamics in drug discovery (2024) · doi

  AlphaFold 3 claim in their publication that the performance of their models on protein-ligand systems is better than classical docking tools, such as Vina97,98, and greatly outperforms all other blind dockings like RoseTTA- Fold All-Atom. Their evaluation was done on their PoseBusters benchmark set, which is composed of 428 protein-ligand structures that were not included in their training. The reported accuracy, as the percentage of protein-ligand pairs with pocket-aligned ligand root mean squared devia- tion of less than 2 Å, is over 90% for their high-confidence group. AlphaFold and other Machine learning models may be limited in the near future in their ability to predict cryptic pockets, given limitations in training data. It may be possible to create training data by enhanced sam- pling MD and use this data to train machine learning models to produce additional target structures. Related research is in Lyu, etc.99, where the authors followed up on several hundred computational hits and found that there was little to no overlap for the same receptor when starting with the AlphaFold 2 model versus the experimental structure. This indicated that AlphaFold models are already showing some potential on modeling dif- ferent conformations.

  Keywords: alphafold models ligand protein training structures machine learning claim publication performance systems better classical docking
• Assessing the Generalizability of Machine Learning and Physics-Based Methods with DNA-Encoded Libraries (2026) · doi

  Predicting protein-ligand binding is a central challenge in computational drug discovery, and while machine learning (ML) and co-folding methods have advanced rapidly, their ability to generalize beyond training or parameterization regimes remains insufficiently understood.

  Keywords: predicting protein ligand binding central challenge computational drug discovery machine learning folding advanced rapidly ability
• Towards Generalizable Protein-ligand Co-folding with ACER (2026) · doi

  While recent co-folding models such as AlphaFold-3 achieve accurate structure prediction, they fail to generalize to underexplored binding interfaces - systematically misplacing ligands, particularly for allosteric or structurally novel targets.

  Keywords: recent folding models alphafold achieve accurate structure prediction fail generalize underexplored binding interfaces systematically misplacing