Docking fails for d-limonene in OR1A1.

[electronicsbyjulie]

The best-binding algorithm always gives C5 - PHE206 as the best binding pair, but that part of the binding pocket doesn't have enough space for the ligand, causing it to clash with various residues and preventing finding valid poses. There's more space near TYR258 for the ligand.

Secondarily, it is identifying C1 - MET255 as the second pair. This is a vdW interaction between aliphatic atoms, and altogether ignores C7 which is a pi atom.

A better bind would be C5 - TYR258 and C7 - PHE206 or TYR251. Then C9 could pair with either MET255 or ILE105 so that the ligand would sit vertically inside the pocket, not flat against the PHE206 ring trying to straddle a space that's too small for it to fit.

[electronicsbyjulie]

1. Residues listed in the RES option of CEN and PATH options should be given equally the highest weight for best-bind functionality. TYR251 and TYR258 are important GPCR activation residues so for a known agonist like d-limonene the docker should attempt to bind them to the ligand if no stronger potential binding exists nearby.
2. The RES option should be preferred whenever important receptor activation residues are known. Users should not assign docking residues willy nilly, but should be cognizant of the algorithm placing extra attention to the indicated residues.
3. Ultimately, the best-binding algorithm should influence poses, not determine them. Molecules are going to do whatever gets them into the lowest energy configuration, and won't always find a GPCR activating pose if an inactive configuration is a better fit.

[electronicsbyjulie]

Visually, it looks like PHE206 and TYR251 should both point "down" towards the cytoplasmic direction, with either C5 or C7 in the space between them, with the other ligand pi bond stacked with TYR258 which would also point "down". But the current algorithm does not find this configuration.

[electronicsbyjulie]

Possible solution to the problem:

1. For each residue within about 10A of the binding center, whose side chain does not face away from the center, determine its outer sphere within which ideal binding is possible, and its inner sphere within which ideal binding is impossible because the atoms would be too close together. Ideal binding means interacting atoms are at their ideal distance for the type of interaction, e.g. H-bond, pi stack, etc.
2. Do full rotations in 30 degree increments of the ligand about the X, Y, and Z axes. So for example this could mean do a full rotation about X, then go 30deg about Y, repeat X, so on until Y finishes, then do 30deg about Z, etc until Z finishes. That gives you a tumble that exhaustively examines about 1700 potential starting poses.
3. During this exhaustive tumble, record which XYZ axis rotation angles give the best sum of ligand atoms inside the outer spheres of compatible residues (so e.g. a methyl group and an isoleucine, an aldehyde and a serine, etc), giving extra weight to binding residues from a CEN RES or PATH RES parameter.
4. It may help to jostle the ligand slightly within the exhaustive tumble, perhaps trying random 0.5A motions to see if any of them improve the score for that XYZ rotation.
5. After identifying the best rotational configuration, do full rotations on the involved residues' bonds to minimize clashes.
6. Once all this is done, go through the normal iterations just to fine tune the dock.

Since this is a somewhat involved task, it's best to do it as a PR.