Automatic Tautomer Generation

We have developed a physics-based method of tautomer generation uses predicted physical properties of the molecule in aqueous solution rather than predefined rules for protonation. The steps of this approach can be summarized as the following:

1. Rapidly evaluate the micro-pK_a associated with a protonation/deprotonation process at each atom for the initial molecule. We use a similarity-based micro-pK_a predictor, which compares the process to a training set of reference data.

2. The predicted micro-pK_as are used to categorize each atom as "active" or "inactive". An "active" heavy atom is one that is most likely to be a protonation/deprotonation site, while an "active" hydrogen is one that is involved in the proton transfer. The protonation states of "active" atoms are also taken into account.

3. Tautomers are generated by a combinatorial permutation of all the active atoms that have been identified.

We believe that any errors stemming from the micro-pK_a predictor will have little effect on the final result, since they will most likely involve weak proton donors and acceptors with high energy conformers which are filtered out in the Macro-pKa workflow. The details of the algorithm have been published elsewhere [312]