Big bucks — but what's the payoff?
The Stanley Medical Research Institute, a Maryland-based philanthropy, is donating $100 million to uncover the genes important in mental illnesses such as bipolar disorder and schizophrenia, according to an article in today's Boston Globe.
The money is going to the Broad Institute, led by genome bigwig Eric Lander, who was one of the driving forces behind the cancer genome. I've already noted the criticisms against that project, and some of the same apply here. Sure, technology now allows us to find the genetic variations between different people and the researchers will no doubt find masses of data.
But these are extremely complex disorders, each involving multiple genes. What roles do those genes play in the disease? Without understanding how the different genes interact and what the impact is of the different variations, the data will be all but meaningless. For example, scientists from the cancer genome project are reporting in this week's Nature that the number of mutations that drive cancer is much larger than they expected.
To the institute's credit, the mental illness project's results will be publicly available — the more scientists who can analyze the data the better. Lander is quoted in the Globe as saying, "If you're looking for a needle in a haystack, and you can sift the whole haystack, you'll find the needle."
Hmmm.... I don't think that was the message of the idiom.
Comments
There is a slight "problem" with the notion that more data will "poof" bring the answer. These are extremely complex systems with very large numbers of coupled non-linear parameters. Their behavior is inherently chaotic.
Coupled non-linear systems of even a few variables (like 5), are completely intractable to model. How can anyone expect to model something with hundreds or thousands? Particularly when most of the parameters are unknown?
I am not saying that such research should not be done, but lets not pretend that understanding the data will be easy, or even of comparable difficulty to getting the data. Understanding the data will be many orders of magnitude more difficult than getting it.
Posted by: Dave Whitlock | March 26, 2007 08:10 AM
Candidate genes? We already have many more candidate genes than we know what to do with. I'd argue that candidate genes are, in fact, part of the "genetic noise" Debra refers to.
What this field needs is, among other things, new animal models that allow us to make more thoughtful experiments on the biological basis of psychiatric disease.
In the absence of good models in which to test the functional relevance of whatever genes the screening process identifies, the massive investment that Apoorva highlighted will bring rather paltry returns.
Posted by: Juan Carlos Lopez | March 12, 2007 04:50 PM
Thanks, Apoorva, for drawing attention to this announcement. I have to disagree with some of your concerns. As you point out, psychiatric and neurological diseases are incredibly complex, which is why datasets of this size are unfortunately necessary to cut through the genetic noise. This month's Nature Genetics has a good example of the utility of such large studies in identifying candidate genes and genetic regions contributing to brain disorders. The link for this article and a longer rant on this topic is available on Action Potential (http://blogs.nature.com/nn/actionpotential/2007/03/i_beg_to_differ.html).
Posted by: Debra Speert | March 12, 2007 10:33 AM