Predicting multiple conformations of ligand binding sites in proteins suggests that AlphaFold2 may remember too much

What is it about?

After the success of protein structure prediction by AlphaFold2 (AF2), recognized by the 2024 Nobel Prize in Chemistry, interests turned toward generating realistic conformational ensembles. It was shown that running AF2 with stochastic subsampling of the multiple sequence alignment (MSA) can generate alternative protein conformations, and hence the approach received substantial attention. However, the method works only for some fraction of the proteins tested, and the origin of this limitation was not understood. This paper explored the opening of cryptic ligand binding sites in 16 proteins, where the closed and open conformations define the expected extreme points of the conformational variation. Due to the many structures of these proteins in the Protein Data Bank (PDB) the authors were able to study whether the distribution of X-ray structures affects the distribution of AF2 models. They have found that AF2 generates both a cluster of open and a cluster of closed models for proteins that have comparable numbers of open and closed structures in the PDB and not too many other conformations. This was observed even with default MSA parameters, thus without further subsampling. In contrast, with the exception of a single protein, AF2 did not yield multiple clusters of conformations for proteins that had imbalanced numbers of open and closed structures in the PDB, or had substantial numbers of other structures. Subsampling improved the results only for a single protein, but very shallow MSA led to incorrect structures.

Featured Image

Why is it important?

Its was shown that predicting multiple conformations requires comparably sized clusters of open and closed structures in the Protein Data Bank, whereas rarely seen conformations are usually not predicted, which emphasizes the limitations of the machine learning based methodology.

Perspectives