Speaker
Description
Accurate estimation of protein–ligand binding affinity is the cornerstone of computer-aided drug design. We have developed a universal physics-based scoring function, named SQM2.20, addressing key terms of binding free energy using semi-empirical quantum-mechanical computational methods. SQM2.20 incorporates the latest methodological advances while remaining computationally efficient even for systems with thousands of atoms (with computation time ~ 30 minutes at one core). [preprint: https://doi.org/10.26434/chemrxiv-2023-zh03k ]
To validate the SQM2.20 scoring function rigorously, we have compiled a benchmark dataset consisting of high-resolution crystal structures and reliable experimental affinities for ten diverse protein targets, the PL-REX data set.
Comparative assessments demonstrate that SQM2.20 outperforms other scoring methods and reaches a level of accuracy similar to much more costly DFT calculations. In the PL-REX dataset, it achieves excellent correlation with experimental data (average R2 = 0.69) and exhibits consistent performance across all targets.
To address the remaining limitations of the SQM methods while keeping our physics-based approach independent of prior data on protein-ligand interactions, we are developing a Δ-ML approach which brings all the calculations in the scoring protocol close to the accuracy of DFT. Our results show that it is possible to achieve excellent results not attainable by either uncorrected SQm calculations or pure ML approach without the SQM component.
