Case Studies

The ProtoDiscovery™ platform has been successfully applied to a diverse range of drug discovery projects, both partnered and in-house. A number of brief summaries relating to recent projects are described below:

Case Study 1 : Hit finding using de novo design with ProtoBuild™

An X-ray structure of a nuclear hormone receptor was used as the start point to identify novel agonists of this target using ProtoBuild. The software approach identified 3 new scaffolds which were deemed to be novel in the field. One scaffold was synthesised and was found to exhibit higher potency than other marketed compounds against the same receptor. The project was completed within 3 months.

Case Study 2 : Lead optimisation via de novo design with ProtoBuild™

An active molecule series had been identified and all obvious SAR had been generated. An x-ray structure was available of the target but no co-crystal solved of the active lead. ProtoBuild™ was used to dock the active compound and design replacements for a key part of the active molecule. Eight compounds were recommended for synthesis; 5 compounds were made and 3 showed greater activity than the original compound and provided new chemotypes for further development.

Case Study 3 : Virtual screening with ProtoScreen™ – enzyme target

Identification of new actives were required against an enzyme target with an available crystal structure. ProtoScreen™ was used to screen ~3 million compounds against the target from which 140 novel compounds were selected for testing. Fourteen of the compounds showed good activity with the best compound showing an affinity of 88nM for the target. The compound is currently being optimised using medicinal chemistry and virtual library generation approach.

Case Study 4 : Virtual screening and lead optimisation with ProtoScreen™ – GPCR target

5-HT1A agonists are currently marketed for anxiety (e.g. Buspirone). There is a requirement to design new 5-HT1A agonists with better selectivity profiles than such market compounds. ProtoScreen™ was used to screen 3 million compounds against an agonist-based GPCR model. Of 51 compounds selected for testing 37 compounds were active at 10µM and 4 novel low potent series identified (<50nM). The compounds were confirmed as agonists in activity assays. The hit finding phase of the project was completed in 3 months. One of the series has now been optimised into a very potent novel and selective 5HT1A agonist.

Case Study 5 : Virtual screening of a protein-protein interaction

A virtual screening campaign was conducted using ProtoScreen™ approach against a protein-protein interaction identified as a novel potential anti-infective target. The screen of 5M compounds resulted in the testing of 86 compounds of which at least 6 showed good activity and some of which were start points for current clinical studies.

Case Study 6 : Virtual library generation with ProtoScreen™

Optimisation of a portion of an existing lead inhibitor was required to overcome IP issues and improve the existing PK profile.

A virtual library approach was undertaken with a synthetically accessible library of ~6,000 compounds enumerated and docked using ProtoScreen™ to the target site. Careful examination of patents and literature was used as an initial filter to flag compounds that were outside existing IP. From the docking scores a shortlist of 100 compounds were marked for synthesis and testing.

All compounds synthesized had activities ranging from µM to nM, giving valuable SAR. A small number of control compounds, not selected, were inactive. This approach allowed the rapid, cost-effective generation of novel compounds and the exploration of compounds with alternative PK properties to the lead. It was subsequently found that the predicted pose generated with ProtoScreen™ matched the actual pose of the inhibitor in a proprietary co-crystal structure.

Case Study 7 : 2D virtual screening with ProtoClassify™

ProtoClassify™ was used to create a model able to discriminate between a set of known actives and random molecules for an aminergic GPCR target. The model was used to identify 11 candidates for testing from a library of ~3 million compounds [PCD]. Of these compounds 6 were confirmed as active on testing with the best compound showing an activity of 80nM.


©Prosarix 2007-12 ¦ Newton Hall, Town Street, Newton, Cambridge, CB22 7ZE ¦ +44 (0)1223 870400 ¦ ¦ legal