Are Deep Learning Structural Models Sufficiently Accurate for Free Energy Calculations? Application of FEP+ to AlphaFold2 Predicted Structures

Abstract

The availability of AlphaFold2 has led to great excitement in the scientific community - particularly among drug hunters - due to the ability of the algorithm to predict protein structures with high accuracy. However, beyond globally accurate protein structure prediction, it remains to be determined whether ligand binding sites are predicted with sufficient accuracy in these structures to be useful in supporting computationally driven drug discovery programs. We explored this question by performing free energy perturbation (FEP) calculations on a set of well-studied protein-ligand complexes, where AlphaFold2 predictions were performed by removing all templates with >30% identity to the target protein from the training set. We observed that in most cases, the ∆∆G values for ligand transformations calculated with FEP, using these prospective AlphaFold2 structures, were comparable in accuracy to the corresponding calculations previously carried out using X-ray structures. We conclude that under the right circumstances, AlphaFold2 modeled structures are accurate enough to be used by physics-based methods such as FEP, in typical lead optimization stages of a drug discovery program.

Benchmarking Refined and Unrefined AlphaFold2 Structures for Hit Discovery

“…Chen et al performed short MD simulations prior to virtual screening with docking to identify potential WSB1 inhibitors. Ray and co-workers explored the retrospective FEP performance on AF2 models, using Schrödinger’s FEP+ package, together with a popular benchmark data set for evaluating the performance of FEP implementations. The authors modified the AF2 inference protocol to withhold templates and sequences of proteins with sequence identity above a threshold when generating a predicted structure.…”

Are Deep Learning Structural Models Sufficiently Accurate for Virtual Screening? Application of Docking Algorithms to AlphaFold2 Predicted Structures

“…22 For example, in a recent publication, we showed how a stateof-the-art implementation of FEP (FEP+ from Schrodinger), when applied to AF2 structures, could produce analogous results to those when using crystal structures. 23 In that study, in order to impose more realistic prospective conditions on our benchmark experiment, we developed a custom AF2 version, named AF2 30 , where we eliminated all structural templates with >30% identity to the target protein to make the structure prediction. We observed that, in most cases, ΔΔG values from AF2 30 structures were comparable in accuracy to the corresponding calculations previously carried out using X-ray structures.…”

“…Our AF2 customization has been described in detail in our recent FEP study. 23 Briefly, we systematically removed all template structures above 30% sequence identity from the database used to build the models. Our version is now capable of removing either structural For those cases where there was no pose returned by Glide or where the rmsd was above 2 Å, we provide a short additional explanation.…”

Are Deep Learning Structural Models Sufficiently Accurate for Virtual Screening? Application of Docking Algorithms to AlphaFold2 Predicted Structures

“…Our AF2 customization has been described in detail in our recent FEP study [23]. Briefly we systematically removed all template structures above 30% sequence identity from the database used to build the models.…”

“…For example, in a recent publication we showed how a state of the art implementation of FEP (FEP+ from Schrödinger), when applied to AF2 structures, could produce analogous results to those when using crystal structures [23]. In that study, in order to impose more realistic prospective conditions on our benchmark experiment, we developed a custom AF2 version, named AF2 30 , where we eliminated all structural templates with >30% identity to the target protein from the training set.…”