VM2: Computation of Binding Free Energies for Ligand Series Ranking

The VM2 method, developed by VeraChem, calculates true free energies of binding using the mining minima method – a rigorous statistical mechanics based approach. VM2 can provide accurate protein-ligand binding affinities, and recent advances, such as parallel processor implementations for CPU and GPU clusters, as well as ports to Cloud platforms, now allow routine use of VM2 for rank-ordering of ligand series in pharmaceutical industry drug development settings.

Conventional drug lead optimization requires multiple rounds of intensive chemical synthesis and testing to make improvements in properties such as solubility while maintaining potency. Accurate predictions of protein-ligand binding affinities, therefore, provide the potential for tremendous savings with respect to cost and time in the drug development process.

If you think VM2 can help in your drug development efforts contact us at vc@verachem.com or submit a request to get VM2 below.

VM2: Fast, Accurate Protein-Ligand Binding Affinity Predictions

Benchmarking of VM2 compared to experiment shows fast, accurate prediction of protein-ligand binding affinities. The following is a comparison of VM2 Pearson correlation with experiment R with FEP+ and PMX/GROMACS for the Gapsys benchmark set:

VM2 correlation with experiment R compared with that from FEP+, Glide Dock, Glide Score, and Prime (MM/GBSA) for the Merck benchmark set:

VM2 wall time per ligand is 20-30 minutes on 12 CPUs/7-10 minutes on CPUs/GPUs. As VM2 calculations are entirely independent, given a modest sized CPU cluster, each entire benchmark can be run in several hours total. See this video presentation for further details on these VM2 benchmarks.

Additional VM2 Performance Examples

These plots are of VM2 calculated binding free energy predictions versus experimental binding affinities for the protein receptors P38A kinase, HIV-1 protease, PDE10A, and BRCA1, and corresponding inhibitor series ranging between 16 and 38 ligands. These data were generated both by VeraChem and academic users. An all-atom AMBER force field was used in each case with PBSA corrected solvation energies.

Results generated by  industry partners for two protein receptors, one with a series of 170 ligands; the other with a seres of 50. An all-atom OPLS force field with PBSA corrected solvation energies was used.