The state you dock matters: protonation and tautomers

Most docking failures are blamed on the scoring function. A surprising number are actually ligand-preparation failures: you docked a molecule in a protonation state or tautomer that does not exist at physiological pH, and no scoring function can rescue a chemically wrong input. The SMILES you drew is not the species that sits in the pocket. Getting from one to the other is the quiet, decisive step that happens before any pose is ever scored.

Why the drawn structure is not the docked structure

A SMILES string encodes one fixed arrangement of atoms, bonds, and formal charges. In water at pH 7.4, a real drug-like molecule is an equilibrium mixture of microspecies: protonated and deprotonated forms of every ionizable group, plus tautomers that shuttle a hydrogen and a double bond between two positions. A basic amine is mostly protonated and positively charged. A carboxylic acid is mostly deprotonated and negative. An imidazole or a 2-hydroxypyridine flips between tautomers that present hydrogen-bond donors and acceptors in completely different places. Each of these changes the molecule's net charge, its hydrogen-bonding pattern, and therefore the interactions a docking engine can score.

How much does it actually change the result?

A lot. Brink and Exner's systematic study fed different protonation, tautomeric, and stereoisomeric states of the same ligands into GOLD and PLANTS and found that both programs frequently failed to identify the correct protomer for a given complex, and that the chosen state materially changed both the predicted pose and the score. The practical lessons:

• Charge drives placement. Flip an amine from neutral to protonated and you add a salt-bridge or strong hydrogen bond that can reorient the entire pose toward an acidic residue. Dock the neutral form by accident and you miss the interaction the molecule actually makes.
• Tautomers move the donors and acceptors. The same heteroaromatic ring can read as a donor in one tautomer and an acceptor in another at the same atom. Pick the wrong one and the hinge hydrogen bond that defines a kinase inhibitor's binding mode simply will not form.
• The lowest-energy state in water is not always the bound state. The pocket can stabilize a minor microspecies. This is why best practice is to enumerate and dock an ensemble of highly-populated states rather than committing to a single one.

What ligand-preparation tools do

The job of a preparation step is to turn one drawn structure into the set of physically relevant 3D species. pKa predictors such as Epik extend the classic Hammett and Taft schemes to estimate micro-pKa values, enumerate the protonation states populated at a target pH, rank them by energy, and attach a penalty to states that are unfavorable under physiological conditions. Tautomer enumerators do the analogous job for hydrogen/double-bond shifts. The output is not a single molecule but a small ensemble, each member tagged with how likely it is to exist. You dock the ensemble and let the scoring compare like-for-like.

A practical default

For a routine run: generate states at pH 7.4, keep every microspecies above a few percent population, and dock them all. If two protonation states score within noise of each other, report both rather than pretending the cleaner number is the truth. The discipline matters most when you are comparing a congeneric series, because a silent inconsistency in how two analogs were protonated will masquerade as a real potency difference.

Try the docking yourself

The fastest way to build intuition is to dock two protonation states of the same ligand and watch the pose and score move. Open Studio and run molecular docking on a basic-amine ligand against a target with an acidic residue in the pocket, then compare the neutral and protonated forms side by side. Liganx is molecular docking online, free and browser-based, so you can test how sensitive your result is to ligand preparation before you trust a single number.

Primary sources

• ten Brink T, Exner TE. Influence of Protonation, Tautomeric, and Stereoisomeric States on Protein-Ligand Docking Results. J Chem Inf Model 49, 1535-1546 (2009). doi:10.1021/ci800420z
• ten Brink T, Exner TE. pKa based protonation states and microspecies for protein-ligand docking. J Comput Aided Mol Des 24, 935-942 (2010). doi:10.1007/s10822-010-9385-x
• Shelley JC, et al. Epik: a software program for pKa prediction and protonation state generation for drug-like molecules. J Comput Aided Mol Des 21, 681-691 (2007). doi:10.1007/s10822-007-9133-z