6
$\begingroup$

I'm interested in predicting the protonation and tautomer state of organic molecules (typically CHNOS) at near pH 7 in water. I've used tools like Schrodinger's Epik / LigPrep before, but they definitely make wrong predictions sometimes, and I don't know exactly how they work under the hood. I'd like to try to do something based on running QM calculations myself but I'm really not familiar enough with this particular subfield to know what is considered the best approach.

What is the current state of the art method for predicting protonation and tautomer state?

P.S. Related: what is the standard benchmark for this?

$\endgroup$
1
  • 3
    $\begingroup$ Since you mention interest in tautomer state of molecules, tautobase (J. Chem. Inf. Model. 2020, 60, 1085–1089; doi 10.1021/acs.jcim.0c00035) with public searchable pairs of tautomers (GitHub, includes logK data) may be an interesting reference to check if the preferred tautomer by your computation is the one experimentally observed. $\endgroup$
    – Buttonwood
    Jan 13, 2022 at 21:20

1 Answer 1

6
$\begingroup$

While I don't know how particular protonation / tautomer tools work, most are trained with various ML models to predict pKa in water. Fortunately, we have a lot of experimental data for that task, although sometimes of varying quality.

For pKa prediction, there are a few public databases:

I'm aware of two major open source models:

You mentioned an interest in quantum calculations for pKa. This would approach the proton affinity.

To my knowledge, the current state of the art is from the Grimme group: "High accuracy quantum-chemistry-based calculation and blind prediction of macroscopic pKa values in the context of the SAMPL6 challenge"

Their CREST / xtb tool can do semi-automated protonation, deprotonation, and tautomer (deprotonation followed by protonation) screening using the GFN2 semiempirical DFTB method. Generally they recommend following the GFN2 energy calculations with higher-level DFT calculations.

There are similar papers by other groups (i.e., fast semiempirical calculation, followed by DFT energy calculations of the different protonated or deprotonated forms).

$\endgroup$
1
  • 1
    $\begingroup$ If you're interested in combining ML and QM, drop me a line. We have some promising results in the works. $\endgroup$ Jan 14, 2022 at 19:10

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .