THINK docking and scoring of fragments (to 13 May2020)

We have now completed our initial docking and scoring of the submitted molecules using our THINK software which is probably best known for the distributed computing docking projects at Oxford University and Find-a-Drug some 20 years ago.

We used the 1093 (5RF3) crystal structure with 3 centre pharmacophore docking targeting residues that have strong interactions in the non-covalent crystal complexes: (41),(44),(140),(142),(143),(144),(163),(166),(189).

The molecules were filtered using THINK’s standard drug-like criteria which include molecular weight and flexibility but not CLogP.

I’m currently getting a Server Error 500 when trying to valid the SD file for upload but here’s a link to the unhacked and hacked SDF (which attempts to be ver_1.2 compatible).
Highest ranked
Reformatted Top 200

The scoring function is essentially the ChemScore function developed in the 1990s by Proteus with adaptions which eliminates the need for hydrogens and specific tautomers.

1 Like

Hi. I’m not sure, whether there has been some error in docking process, but the structures in the “Highest ranked” file do not quite seem to be low energy conformers as there are multiple C-H C-H bond clashes in the first structures that I saw. Have I misunderstood something or hydrogens have been added post-docking or no conformer search has been made?

Thanks for taking a look and for your feedback. It looks like the conformers generated were refined with an all-atom (rather than explicit H) conformational energy. I’m not sure why this was the case as it’s not a good option. (I had viewed random conformers during the docking and should have taken the time to eye ball these).

Obviously, some of the hits dont have bad contacts and others can be avoided by relatively small conformer changes. However, such changes combined with including explicit Hs will mean that the scores will change - none will improve!

I’m currently redocking these molecules and will upload a revised set of top hits (on Wednesday 20 May). Some of these molecules are really promiscuous (lots of heteroatoms and flexibility) but I have a few processors working on them!

A single conformer of ranked docked molecules is available in SD file format (reformatted for upload). Some of these molecules are crowded and especially these have some contacts which can usually be relieved by a full geometry optimisation or in some cases rotating a terminal group such as a methyl. However, these conformers were saved prior to side-chain relaxation which is normally used with a rotations/relaxations of terminal groups. (Otherwise a vast number of corresponding PDB files would be needed). Water molecules in the binding site are ignored by the scoring function if they conflict with the docked molecule.

My selection of a top 200 excludes those with bad contacts in the selected conformer.

I have spent hours yesterday and the day before attempting to upload these sd files but always get a SERVER ERROR 500. I’m still unclear whether this is a genuine server error, a browser related problem (eg the target page not being compliant to HTML 5), something to do with sd file extensions that we use or some obscure aspect of the sd file variant required which I’ve not recognised. I’ve tried every variant that I can remember encountering over the last 35 years without success!

Many thanks for taking the time and effort to do the docking and generate your selection!
I’ll pass your problems about file upload onto Rachael who has been the person working hard to address these issues.
Your selection will now be added to the other selections made using alternative computational methods and we’ll keep you updated on progress.