Prediction of the binding affinity of compounds with diverse scaffolds by MP-CAFEE

- We predicted binding affinities of six compounds for p38α MAP kinase by MP-CAFEE.
- The compounds have diverse scaffolds and published X-ray co-crystal structures.
- Predicted and experimental binding free energies correlate well (R2 = 0.93).
- We could rank the compounds with different scaffolds using MP-CAFEE.
- We proposed the optimal sample sizes to identify or optimize lead compounds.

Accurate methods to predict the binding affinities of compounds for target molecules are powerful tools in structure-based drug design (SBDD). A recently developed method called massively parallel computation of absolute binding free energy with a well-equilibrated system (MP-CAFEE) successfully predicted the binding affinities of compounds with relatively similar scaffolds. We investigate the applicability of MP-CAFEE for predicting the affinity of compounds having more diverse scaffolds for the target p38α, a mitogen-activated protein kinase. The calculated and experimental binding affinities correlate well, showing that MP-CAFEE can accurately rank the compounds with diverse scaffolds. We propose a method to determine the optimal number of sampling runs with respect to a predefined level of accuracy, which is established according to the stage in the SBDD process being considered. The optimal number of sampling runs for two key stages-lead identification and lead optimization-is estimated to be five and eight or more, respectively, in our model system using Cochrans sample size formula.