October 30, 2003

Beeswax

Sridhar Ramaswamy was just down for a visit and gave a very nice presentation of his work. Over dinner and in his talk the following day, he's taken the view that immediate clinical applicability and pure scientific curiousity are at odds when it comes genomics.

Essentially, he suggests, the key is to reconcile long-term outcome with gene expression patterns, and this is only possible with formalin-fixed tissues in paraffin blocks because adequately prepared fresh frozen tissue samples are too "young" to yield us long-term follow-up data. Therefore, understanding that paraffin blocks will yield far few gene expression signals, why not identify representative genes that lie near the centroids of broader signatures and RT-PCR these signals out of tissue blocks?

This is a compelling argument, and I'd have to counter this way:
1. Broadly, to really understand cancer requires basic science moving hand-in-hand with clinical insight; and this can only happen with frozen tissues and microarray analysis.

2. More specifically, the key difference between what our group is doing and what other groups are doing is not to stop at one classification, but to keep on going, and whittling the complex pathophysiologic phenomenon of cancer down to the individual. Keep on subdividing your data until you can't divide any further, and what you have left is the individual patient with all the idiosyncratic characteristics of her disease. You can't do this with 70 genes. You might be able to make the first cut, but the information peters out after this.

3. Banking frozen tissue, and more importantly, banking all the information you cull from it is an investment for the future. There is little chance that your initial analytic approaches will properly mine all the ore out of your data. Why go through the trouble of creating massively multivariate data and throw away most of it?

Posted by erich at 08:31 PM | Comments (0) | TrackBack

October 08, 2003

Next Steps

With the recent spate of publications comes thinking about the next phase of the Project. There are several developments that it is premature to talk about, but it's clear now that every group interested in bringing genomics to day-to-day clinical application have to begin studies involving many hundreds to thousands of patients. Only with larger and more heterogeneous populations of patients can we test whether promising trends represent biological reality.

More to come.

Posted by erich at 05:56 PM | Comments (0) | TrackBack