Estimating entropy is crucial for understanding and modifying biological systems, such as protein-ligand binding. Current computational methods to estimate entropy require extensive, or at times prohibitively extensive, computational resources. This article presents SHAPE (SHape-based Accurate Predictor of Entropy), a new method that estimates the gas-phase entropy of small molecules purely from their surface geometry. The gas-phase entropy of small molecules can be computed in ≈ 0.01 CPU hours with an average run time of O(√Na), where Na is the number of atoms. The accuracy of SHAPE is within 1−2% of computationally expensive quantum mechanical or molecular mechanical calculations. We further show that the inclusion of gas-phase entropy, estimated using SHAPE, improves the rank-order correlation between binding affinity and binding score from 0.18 to 0.40. The speed and accuracy of SHAPE make it well-suited for inclusion in virtual screening applications.

No abstract