pKa Prediction with Graph Neural Networks

Epik7 is a new machine learning model for predicting the pKa values and protonation states of complex, drug-like molecules. It is trained on over 42,000 pKa values from a variety of sources, and can accurately predict pKa values with median absolute and RMS errors of 0.42 and 0.72 pKa units. Epik7 can also generate protonation states, and recovers 95% of the most populated protonation states compared to previous versions.

Epik7 can also be used to evaluate protonation states for crucial molecules and prepare ultra-large libraries of compounds to explore vast regions of chemical space.

Check out the full work here