What docking actually computes
Molecular docking answers two questions: in what orientation and shape (the pose) does a ligand fit a pocket, and how good is that fit? A docking program does this in two stages. First, a search explores many positions, rotations, and internal twists of the ligand inside the pocket — which is why conformer generation and bond rotation matter so much. Second, a scoring function ranks each candidate pose with a quick estimate of binding strength.
Good poses obey the chemistry you learned in the first guide: they show shape complementarity with the pocket and place donors against acceptors. Most standard docking treats the protein as rigid, which is fast but ignores induced fit — the way a pocket can reshape to accept a ligand. When the real pocket flexes, rigid docking can miss the right answer entirely.
Why scoring is fast but rough
A scoring function must evaluate millions of poses, so it trades rigor for speed. It sums simple terms — hydrogen bonds, hydrophobic contact, clashes, sometimes a crude desolvation term — into a single number. That number correlates loosely with binding, but it routinely gets the ranking wrong by a couple of orders of magnitude in affinity. Docking is good at finding a plausible pose and at separating clear binders from clear non-binders; it is poor at telling you that compound A is 3-fold better than compound B.
Docking for virtual screening
The biggest payoff is virtual screening: docking a library of thousands or millions of molecules and keeping the top-ranked ones to test experimentally. Treat the score as an enrichment tool, not a verdict — it is meant to give you a hit-rich shortlist, not to rank that shortlist precisely. Expect many false positives: poses that score well for the wrong reasons, strained conformers, or molecules that would never be soluble.
- Prepare the protein: pick a relevant structure, add hydrogens, set protonation/tautomer states, and decide which waters to keep.
- Prepare ligands: generate reasonable conformers, correct protonation, and remove implausible structures.
- Re-dock the native ligand to validate the workflow before screening anything new.
- Dock the library, then have a chemist eyeball the top poses — never advance a hit you have not looked at.