GeinDock Suite
Home › Flexible docking

Flexible docking

Letting selected receptor side chains move during the search: how to choose them, what the torsion budget costs you, and how to tell whether the flexibility helped.

A rigid receptor is a useful approximation, and it is wrong in a specific and predictable way: binding sites are not rigid. A side chain that would rotate aside to admit a compound is, in a rigid receptor, an immovable wall, and the search reports that the compound does not fit. Flexible docking removes that wall for a small number of residues you choose.

Select Docking › Docking Mode › Flexible Docking. Step 4 becomes Flexible Side Chains.

A flexible run with no flexible residue is a rigid run wearing the wrong label. This is why the run is blocked until at least one side chain is chosen, rather than quietly falling back. If it ran anyway you would have a result labelled flexible that was produced by a rigid search, and nothing in the output would tell you.

The side chain picker

Choose Side Chains opens the picker. It has already made a selection for you when it opens, using the binding site as the reference.

The flexible side chain selection dialog
The picker on opening. Eleven side chains fell within the cutoff, six were kept for a total of fourteen torsions, and five were skipped to keep the search space bounded.

Automatic selection

Three numbers drive the automatic choice, and all three are shown so that the selection is explicable rather than magic.

Control Default What it does
Contact cutoff4.5 ÅHow close a side chain must come to the reference ligand or the box to be a candidate. This is the distance at which a residue is genuinely lining the pocket rather than sitting behind it.
Max residues6An upper bound on how many are kept, however many qualify. Beyond a handful, the search space grows faster than the benefit.
Torsion budget14An upper bound on the total rotatable bonds contributed by the selection. This is the number that really matters, because torsions and not residues are what the search has to explore.

Auto-select re-runs the selection with the current numbers, and Clear empties it. The summary line at the top of the dialog reports what happened and why, so you can see both what qualified and what was left out.

The side chain table

Column What it tells you
Chain, Residue, SeqWhich residue this is.
To ligand (Å)How close its side chain comes to the reference. The table is sorted by this, so the residues most likely to matter are at the top.
TorsionsHow many rotatable bonds this residue would contribute. Alanine and glycine show n/a, because they have no side chain torsions to offer and making them flexible would achieve nothing.
FlexibleThe tick box. This is what you change.
Residues with no side chain torsions cannot be made flexible. Glycine has no side chain at all and alanine has only a methyl group. Both appear in the table when they are close to the site, so you can see that they were considered, but neither can be selected.

Choosing by hand

Three ways to override the automatic choice.

The residue selection sub dialog
The residue picker. Every candidate carries its torsion count and its distance, so a selection made by hand is as informed as the automatic one.

The Receptor selector box at the foot of the dialog always shows the current selection as a text expression, and Apply text reads one back in. That is the field to copy into your notes, and the field to paste into when a colleague sends you a residue list.

Only within the cutoff above the table hides the residues that did not qualify, which makes the list shorter when you are working through a crowded site.

The torsion budget, and why it is the number to watch

The footer reports the running total, in the form four side chains, ten of fourteen torsions, and warns that making a side chain flexible enlarges the search and can introduce false positives.

Each flexible torsion adds a dimension the search must cover, exactly as a rotatable bond on the ligand does. Six flexible side chains contributing fourteen torsions makes the problem roughly as hard as adding fourteen rotatable bonds to your compound. At a fixed exhaustiveness that means the search covers the enlarged space more thinly, so a flexible run that is not given more search effort can easily be worse than the rigid run it replaced.

Flexibility manufactures false positives if you overuse it. A pocket with many mobile side chains can accommodate almost anything, because the search will deform the site to fit whatever it is given. That produces good scores for compounds that do not bind. Make flexible only the residues you have a specific reason to expect to move, and prefer three well chosen ones to six speculative ones.

Reading the result

Poses come back in the usual table, with one column that only this mode fills in.

The pose table with the side chain shift column
The side chain shift column reports, for each pose, how far the flexible residues actually moved from where the input structure put them.

Read that column carefully, because it tells you whether the flexibility earned its cost.

Shift What it means What to do
Small, well under an ÅngströmThe side chains barely moved, so the rigid receptor was already adequate.Go back to ligand docking and spend the search effort on exhaustiveness instead.
Moderate, on one or two residuesA specific side chain rotated to let the compound in. This is the case the mode exists for.Keep it, and check that the moved residue is one you had a reason to expect would move.
Large, across several residuesThe search may have rearranged the pocket into a shape that has no physical basis in order to accommodate the compound.Treat with suspicion. Reduce the selection to the residues you can justify and run again.

The interaction tools work on a flexible pose exactly as they do on a rigid one, and they are the quickest way to see what the movement bought you: a contact that appears only in the flexible run is the concrete argument for having allowed the movement.

A flexible docking result with its contacts listed
A flexible pose with its full contact list, including hydrogen bonds to residues that moved.
The two dimensional diagram of a flexible pose
The same pose as a diagram, which makes the residues involved easy to name in a report.
Compare the flexible run against the rigid one rather than reading it alone. Dock the same compounds both ways. If the flexible run improves the score of compounds you know to be active and does not improve the ones you know to be inactive, the flexibility is modelling something real. If it improves everything equally, it has simply made the pocket more accommodating and told you nothing.