The maximal and current accuracy of rigorous protein-ligand binding free energy calculations

Abstract

Computational techniques can speed up the identification of hits and accelerate the development of candidate molecules for drug discovery. Among techniques for predicting relative binding affinities, the most consistently accurate is free energy perturbation (FEP), a class of rigorous physics-based methods. However, uncertainty remains about how accurate FEP is and can ever be. Here, we present what we believe to be the largest publicly available dataset of proteins and congeneric series of small molecules, and assess the accuracy of the leading FEP workflow. To ascertain the limit of achievable accuracy, we also survey the reproducibility of experimental relative affinity measurements. We find a wide variability in experimental accuracy and a correspondence between binding and functional assays. When careful preparation of protein and ligand structures is undertaken, FEP can achieve accuracy comparable to experimental reproducibility. Throughout, we highlight reliable protocols that can help maximize the accuracy of FEP in prospective studies.

Robust prediction of relative binding energies for protein-protein complex mutations using free energy perturbation calculations

“…These results are consistent with previous protein FEP+ results where accuracy has been at or near the 1 kcal/mol level[17, 18, 36, 35, 39, 40]. This level of accuracy is approaching the intrinsic uncertainty expected from experimental results, which has been proposed to be in the range of 0.4-0.9 kcal/mol for protein-protein and protein-ligand SPR binding measurements [34, 48, 39].…”

“…These results are consistent with previous protein FEP+ results where accuracy has been at or near the 1 kcal/mol level [17,18,36,35,39,40]. This level of accuracy is approaching the intrinsic uncertainty expected from experimental results, which has been proposed to be in the range of 0.4-0.9 kcal/mol for protein-protein and protein-ligand SPR binding measurements [34,48,39]. Some of the largest errors in the full benchmark dataset arose for cases where conformational rearrangement was either observed (as seen with the change in conformation of 1DVF wt residue D:H33) or expected (e.g.…”

A foundation model for bioactivity prediction using pairwise meta-learning

“…More importantly, we found that by using 40% of the data for fine-tuning (12 compounds on average for each assay), our model can surpass the performance of FEP calculation tool FEP+(OPLS4), 37 which is the latest release of the leading commercial software developed by Schrödinger using state-of-the-art OPLS4 force field. 11, 43 We also compared with the state-of-the-art structure-based methods PBCNet 19 following their experimental settings, and we found that ActFound outperforms PBCNet even without using any given target protein information ( Supplementary Figure 14 ). The promising performance on two FEP benchmarks indicates that our foundation model can serve as a machine-leaning-based alternative to FEP calculations by only requiring a few labeled data points for fine-tuning.…”

“…21 After data leakage processing, 58.3% of the compounds in the FEP set are unseen in ChEMBL. Data for the FEP set assays and the FEP+(OPLS4) 37 calculated results were downloaded from https://github.com/schrodinger/ public_binding_free_energy_benchmark, 43 we directly read the compounds' smiles from the SDF files, and manually deleted all formal charges.…”

Multistate Method to Efficiently Account for Tautomerism and Protonation in Alchemical Free-Energy Calculations

“…In recent years, alchemical free-energy calculations have become established as essential tools to support drug discovery projects by helping to prioritize compounds for synthesis. − The groundwork behind these methods dates back to pioneering theoretical work in the 20th century, − followed by subsequent efficiency improvements relying on their combination with various enhanced-sampling algorithms (for examples, see refs – ). While these methods perform well for series of congeneric ligands, , it is typically challenging to estimate relative binding free energies between molecules with different net charges, scaffolds, or binding modes. , In particular, transformations involving ligands with different charges face two main challenges.…”