ChEMBL

A few people have mailed me about getting hold of information on 'failed' drugs, or drugs in clinical development. This is one of those things that sounds really simple, should be readily available but isn't. However, such a list would be really useful in many applications in life science informatics, for example in understanding or anticipating the chemotypes of drug failure, to provide data for second or third generation agents with differentiated profiles (MeToo's are not all bad you know!), to perform 'drug rescue' studies (i.e. in this context to find a new clinical indication for a drug outside of it's originally anticipated use). There are some interesting commercial options for this type research, even for 'off-patent' drugs that can lead to both commercial gain, and significant patient benefit.

For the case of a drug that has been withdrawn post approval, it is quite trivial to get hold of a list of these, it is however, more difficult to get hold of reasons or data for the withdrawl. There are arguably two fundamental reasons for withdrawl, Safety, or Commercial, and both of these words have potent meaning in the Pharmaceutical Business. Do not even think of getting a clear view on the reasons for failure for compounds in development.

Some time ago we started to analyse affinity data of drugs against targets other than their conventional target, and this led to the collection on an ad hoc basis of clinical development candidates against particular targets or target classes. Doing this on a couple of different systems led to a search strategy that makes, in our view, the collection of a complete list of clinical stage development candidates for deposition in the public domain a possibility. This database we call CandiStore, it attempts to capture....

Chemical structure or sequence (it also covers peptides, monoclonal antibodies, aptamers, etc.)

Trade names

Formal name

Research codes

Molecular target(s)

Mode of action (inihbitor, activator, etc.)

Highest development stage (phase 1 thru 4, with 4 being a launched drug)

Lead development company

CAS number

A few other things primarily involved with database consistency and construction

CandiStore is a work in progress, we have about 11,000 structures in total at the moment, but addition of additional data, and digging into earlier development compounds is patchy at the moment. We currently estimate that CandiStore will contain ca. 45,000 compounds when complete. It is an active part of our curation efforts within ChEMBL, but we would be delighted to have collaborators and contributors to this effort, maybe as part of an 'open source' style project. That way, more people could benefit from the data quicker, but any data you submit would need to be yours to submit! As in all our work at the EMBL-EBI this will be freely available to one and all.

CANDiStore is pretty cool too!

There are, of course, several commercial and proprietary databases that address different aspects of this need. Some of them are truly excellent.