Epik: pKa and Protonation State Prediction through Machine Learning


Epik version 7 is a software program that uses machine learning for predicting the pKa values and protonation state distribution of complex, drug-like molecules. Using an ensemble of atomic graph convolutional neural networks (GCNNs) trained on over 42,000 pKa values across broad chemical space from both experimental and computed origins, the model predicts pKa values with 0.42 and 0.72 log unit median absolute and RMS errors, respectively, across seven test sets. Epik version 7 also generates protonation states and recovers 95% of the most populated protonation states compared to previous versions. Requiring on average only 47 ms per ligand, Epik version 7 is rapid and accurate enough to evaluate protonation states for crucial molecules and prepare ultra-large libraries of compounds to explore vast regions of chemical space. The simplicity of and time required for the training allows for the generation of highly accurate models customized to a program’s specific chemistry.

Version notes

Revisions per reviewers' comments suggestions.


Supplementary material

Supporting Information
Composition and characteristics of the training and validation sets; details on the methods, including the Macro-pKa approach, the effects of varying model layer depth, and the use of a “master” atom; an example speciation report; details on the approach used to obtain Epik Classic and Epik v 7 values for the test sets; additional commentary on the results of the test sets; additional details on the similarity between the training set versus the test sets.
Epik v 7 Results on Test Sets
Archive of the Epik v 7 results in CSV form separated by test set