BioBlondeBabe: Bioinformatics

To cif or not to cif

To cif or not to cif, that is the question...

I spent a disheartening afternoon staring at some of the new extensions of mmCif. mmCif is the data dictionary for the RSCB protein databank. http://www.rcsb.org/pdb/ It was designed by and for Xray crystallographers. Unfortunately, other people are trying to use the data in ways that were unforseen by the committee that designed the data dictionary.

Bad database design seems to be a hallmark of bioinformatics projects. Most biologists don't understand the importance of a good data model, or how to anticipate how the data might be mined or used. People who know how to model data don't have the domain expertise to understand what they're modelling. So, dictionaries like mmCif become entrenched. They're great for the handful of domain experts who want to capture every aspect of their experiments, but painfully hard to use in any sort of development project.

Well, enought ranting. The moral of the story is, if you're a biologist, get to know a databaser before you do your next project. If you're a databaser, talk to a biologist one day - they might just need your help.

Clinical trial software will not speed drug discovery

An Rx for the Pharmaceuticals." Getting drugs to market is a costly process without the guarantee of success. Now a cottage industry is springing up to help the big boys save a little dough and pass FDA muster." By Kristen Philipkoski. [Wired News]

Oh, my. This is an article on Pharsight, a software product that is designed to optimize clinical trials. The idea is to speed drug discovery by optimizing clinical trials The product appears to be a biostatistician in a box. Now, the authors of this article suggest that rational drug design and other modeling tools have sped up early drug discovery. I disagree. There were many more new drugs brought to market before the advent of sophisticated computer tools then there are now. Pharsight may be useful, but its impact on drug discovery will probably be minimal because most companies have statisticians on staff with a level of experience and sophistication than cannot be matched by a piece of software.

in the discovery process clinical trial design is not rate limiting. The real problem with drug discovery is the FDA and consumer expectations, not lack of tools. In today's climate, aspirin and penicillin would not have been approved as drugs. Aspirin has many side effects and penicillin can lead to life-threatening allergic reactions. Today's consumer wants a "magic bullet" - a drug that has no side-effects, no risk, can be taken once a day as a pill, and will completely cure their disease in a matter of days. Promising compounds are constantly being removed from consideration because of potential side effects in animal trials. Drugs that must be delivered by injection or taken more than twice a day are also deprioritized.

Pharsight will not change consumer expectations or the over-protective FDA. It might help drug companies save money, but it will not bring drugs to market any faster.

IBM does bioinformatics

IBM is making a run for the biotech market. They've set up a toolkit, located at http://cbcsrv.watson.ibm.com/Tspd.html. It's not half bad for a free bioinformatics toolkit. As well as providing a server, they are working on some new bioinformatics routines, like pattern discovery and annotation. Yeah, OK. Everyone's doing annotation.

Anyway, it's worth checking out if you haven't seen it before.