All Notebooks | Help | Support | About
Show/Hide Keys
12th December 2012 @ 07:25
Continuing on from the previous post and using the CheMBL substructure search feature the GSK TCAMS Database was filtered to give 17 compounds containing the aryl pyrrole substructure and with accompanying pIC50 data. This was combined with a manually filtered data set of OSM compounds with IC50 data to give this training set.

The structures were prepared using the default settings-
Clean Structure
Ionise at pH7

and conformers were generated using the default settings.

The activity threshold was set to PIC50 = 4.5 (30uM) and sites were created with the following parameters

Find common pharmacophore – Max sites 6
Min sites 5

With the constraints to find only 1 hydrophobe and 2 aromatic rings.

The resulting hypotheses were scored for actives and inactive's using the default settings and then clustered to give 6 hypotheses.

All hypotheses were viewed and manually assessed. The selected map had the features at the following co-ordinates.

0 A 8.855380 -8.074250 1.459800
4 A 4.492750 -7.101520 -0.342651
8 H 11.881400 -4.989210 0.281807
9 R 9.213580 -4.790410 0.496400
10 R 10.737100 -1.309760 -0.741114

The sites were based on the GSK compound 535,023 in this conformer. I've manually produced a .chm file in DS3.5 visualizer for download. This has not been confirmed though - if there are any problems with it can you please let me know.

The actives in the training set can be clustered into 5 groups and all map onto the above hypothesis very well



Viewing these poses and with the new data available for compounds up OSM-S-5 a few conclusions can be made.

Seeing as all 5 clustered active groups map onto this hypothesis it was predicted that OSM-S-103 should be active. It turned out that 103 was only slightly active and from this two things can be suggested. First of all it looks like an ester group at this position is the most favoured ad as such the feature at this site need modified. Maybe a carbothioate at this position could be investigated?

It has also been attempted to map the ester derivative of OSM-S-5 onto this hypothesis with no great success. It could therefore be predicted that this analog will be no more active than OSM-S-103.

Re-mapping the active compounds onto the hypothesis with it limited to only give results when hitting all 5 features.


56 compounds were returned - 33 of which are active. The original input file had 47 active compounds. The actives that did not match are contained within this file.


The inactive compounds can be filtered it with further feature mapping.


- Pyrazoles are totally inactive so can be filtered out.
- A Nitrogen appears not to be accepted in the area of 6.53, -6.28, 1.52 however an oxygen here if favoured.
- A size exclusion sphere should be added in the region of 4.53, -5.9, 2.99

Unfortunately, this pipe-lining or advanced screening is beyond my capabilities with the software I'm using here. This is a great opportunity for someone else to get involved in the project.

Finally, The activity of OSM-S-5 and it's primary amide feature have not been mapped due to flex-abilities in the side chain. More examples from the GSK data may help this process but time constraints have not allowed this at the moment
Attached Files