Quantitative Protein–Ligand Binding Affinity Prediction from Quantum-Mechanical Scoring
- 1Institute of Organic Chemistry and Biochemistry of the CAS, Flemingovo nám. 2, Prague, Czechia, 16000
Abstract
Quantitative prediction of protein–ligand (P–L) binding affinities is generally thought to require extensive configurational sampling, which is computationally demanding. Here, we examine how closely a fundamentally different approach, a quantum-mechanical (QM) description of P–L binding energetics based on a small set of representative structures, can approach the accuracy of MD-based free-energy calculations. Large-scale validation of the SQM2.20 scoring function is carried out on the Public Free Energy Benchmark data set, comprising 88 P–L series with 1,113 ligands in total. To compare directly with the experimental binding free energies (∆G bind ), a scaling factor derived independently in the PL-REX data set was applied, yielding SQM2.20 sc . With refined receptor structures and an extended pool of ligand poses, SQM2.20 sc achieved a weighted pairwise root-mean-square error (RMSE) in ∆∆G of 1.41 kcal/mol, close to the 1.26 kcal/mol reported for FEP+. n a subset of 58 targets and 872 ligands, SQM2.20 sc (RMSE 1.41 kcal/mol) outperformed OpenFE (1.74 kcal/mol). These results demonstrate the transferability of the approach across a wide range of protein targets and ligand chemistries, further strengthening the case for SQM scoring as a computationally efficient tool for applications in computer-aided drug design.
Related
- Comparative Analysis of Basic Machine Learning and Rigorous Free-Energy Calculation on Benchmarking Datasets of Protein-Ligand Binding Affinity
- Global protein-ligand binding affinity profiling via photocatalytic labeling
- EVALUATING MOLECULAR DYNAMICS FRAMES FOR BINDING AFFINITY PREDICTION AND VIRTUAL SCREENING
- Reliability of AI Methods in Drug Discovery: Evaluation of Boltz-2 for Structure and Binding Affinity Prediction
- AlphaDTA: integrating AlphaFold3 embeddings and 3D complex structures for drug–target binding affinity prediction
BibTeX
@misc{fanfrlik2026quantitative,
title = {{Quantitative Protein–Ligand Binding Affinity Prediction from Quantum-Mechanical Scoring}},
author = {Jindřich Fanfrlík and Adam Pecina and Martin Lepšík and Jan Řezáč},
howpublished = {ChemRxiv},
year = {2026},
month = oct,
doi = {10.26434/chemrxiv.15010976/v1},
url = {https://chemrxiv.org/doi/full/10.26434/chemrxiv.15010976/v1},
note = {CC BY-NC-ND},
}
RIS
TY - UNPB
TI - Quantitative Protein–Ligand Binding Affinity Prediction from Quantum-Mechanical Scoring
AU - Jindřich Fanfrlík
AU - Adam Pecina
AU - Martin Lepšík
AU - Jan Řezáč
PY - 2026
DA - 2026/10/23
PB - ChemRxiv
DO - 10.26434/chemrxiv.15010976/v1
UR - https://chemrxiv.org/doi/full/10.26434/chemrxiv.15010976/v1
AB - Quantitative prediction of protein–ligand (P–L) binding affinities is generally thought to require extensive configurational sampling, which is computationally demanding. Here, we examine how closely a fundamentally different approach, a quantum-mechanical (QM) description of P–L binding energetics based on a small set of representative structures, can approach the accuracy of MD-based free-energy calculations. Large-scale validation of the SQM2.20 scoring function is carried out on the Public Free Energy Benchmark data set, comprising 88 P–L series with 1,113 ligands in total. To compare directly with the experimental binding free energies (∆G bind ), a scaling factor derived independently in the PL-REX data set was applied, yielding SQM2.20 sc . With refined receptor structures and an extended pool of ligand poses, SQM2.20 sc achieved a weighted pairwise root-mean-square error (RMSE) in ∆∆G of 1.41 kcal/mol, close to the 1.26 kcal/mol reported for FEP+. n a subset of 58 targets and 872 ligands, SQM2.20 sc (RMSE 1.41 kcal/mol) outperformed OpenFE (1.74 kcal/mol). These results demonstrate the transferability of the approach across a wide range of protein targets and ligand chemistries, further strengthening the case for SQM scoring as a computationally efficient tool for applications in computer-aided drug design.
ER -