Monday, November 23, 2009

De-code me

Well, we return to real science after our soap opera of the last post!

We referred the other day to the story that deCode Genetics, the Icelandic genomics company, has gone into Chapter 11 bankruptcy. It's been coming for some time. As a story in the Times put it,
Whatever business errors deCode may have made, a principal reason for its downfall is scientific — the genetic nature of human disease has turned out to be far more complex than thought.
We certainly have a vested interest in saying that this statement is far from the truth. "than thought" refers to those who believed, wished, or tried to quickly capitalize on the notion, never supported by the body of facts, that common diseases were genetic in the sense of being reliably predicted from individual genotypes. Those who hoped or hyped this tale know that today's state of affairs has been predicted and for correct reasons that have, among other things, to do with evolution, for at least a decade.

But the story is not over. deCode will be taken over by someone, and their founder says their activities will continue. Apparently also, another genetic testing company, 23andMe, has just raised its prices, which has been speculated as indicating financial trouble in Silicon City, too. Whether financial recovery will occur remains to be seen, and these entities may morph to other kinds of analysis or testing to stay in business. But it's not the first time; Celera Genomics, which co-sequenced the human genome, folded some years ago on much the same grounds, that genotypes don't effectively predict disease.

The proponents of the current GWAS (genome-wide association studies)-personalized medicine-and-biobank-laden ethos, that has been so good for the grant, equipment, and recreational genomics business, naturally want to keep the tap open on the promise of Heaven-to-come. Director Francis Collins has nominally staked NIH's future substantially on the basis of this belief. "Salvation's just around the corner" is always how faith-healers have justified passing the hat, and the same can be seen in the quote with which the Times story ends:
“DeCode has been very successful using genome-wide association studies, and among the first to publish many discoveries,” said Dr. David Altshuler, a medical geneticist at the Massachusetts General Hospital. But he expressed optimism that the human genome project would succeed despite deCode’s stumble.
“It would be a mistake to draw any connection between the medical promise of the human genome and the success of a specific company and business model,” he said.
This is right to the extent that no sane person thinks that we'll learn nothing more from genetics! But if the quote accurately portrays what David, who is a first-rate scientist said, he might have been being cagey. Nobody is questioning whether the 'genome project' would 'succeed'. But we're not talking about the genome project here: instead, what's questioned is how usefully genotype data predicts disease.

It is true, depending on your definition of 'success', that GWAS has found genes that contribute to disease. But there is no reason to expect, and every biological and evolutionary reason not to expect that major complex diseases will ever be generally predictable by individual genotypes--and that is what the GWAS main findings have actually shown. Commercial prediction (or, ethically more responsible, within-clinic genetic counseling) services will grow in their ability to predict certain kinds of disorder, like classical single-gene diseases. That will include a subset of common 'complex' diseases. But most risk variants cause only small statistical, environmentally affected, rather than certain risk.

The Times article was again incorrect, and rather naively so, in stating that
Natural selection seems to be much more efficient than expected at ridding the population of dangerous genes, even of those that act well after the age of reproduction.
To the contrary, it was always 'expected' that selection would purge very serious mutations, so that mainly the brand-new, or older but highly recessive (low penetrance) ones would be found at any given time. But selection has little way to purge mutants with long post-reproductive effects.

The Times was right, however, that the aggregate effects of large numbers of rare alleles seem to be the reason why family correlation of risk for many or even most traits is high, implicating genetic factors, but we can't find the genes. There is -- and has long been -- a plethora of evidence of many kinds, from many species, to suggest that this is the case.

The problem was the genome frenzy, driven largely by gold-rush mentality, but also genuine frustration at the difficult of understanding the cause of so many important diseases. That frenzy could not be stopped, and many GWAS and biobank studies are being launched or are committed to that will have similar bleak futures; unfortunately they entail commitments to long-term expense, because vested interests cannot easily be slowed. Country after country has sheepishly followed the wealthier countries, and they're now heavily invested in these costly but questionable studies. They could have known better!

The story is subtly deeper than meets the eye, too. deCode was founded on the idea that Iceland was an isolate founded by a small number of individuals and hence greatly reduced variation so that causal variants could more easily be detected. There were always problems with that logic, but it quickly turned out that the founding population wasn't all that small, relative to the questions at hand. Instead, Iceland did provide a great potential advantage: it was to have a whole-population genealogy, connecting everyone in families back for centuries, with comprehensive medical records in recent generations.

There were apparently lots of political and social problems involved, but the great advantage deCode had, at least in principle, was that a genealogy will generally carry fewer variants than a whole population, and linkage studies in families are more statistically able to detect things than comparisons between cases and controls. That loaded the dice in deCode's favor, and they did find a number of things. But the bottom line was that those findings did not reduce complex traits to simple traits. Just as experiments with crosses between just two inbred animals have shown, even very small populations carry enough variation for complex traits to remain complex.

deCode failed to do what its commercial hyperbole promised even with these advantages. From a scientific point of view, that may be a commercial failure but it's a positive scientific contribution to understanding. That understanding should be what is needed to show convincingly that over-geneticizing common diseases is a mug's game, for reasons that we've known for a long, long time. What else is science all about except to learn and adapt to the realities of Nature?

The best way to slow this train so we can get off and go to more creative approaches is to formally curb funding for these kinds of wholesale genetic studies, and thus force investigators to think and think again about other ways to approach these diseases--and not just by scaling up their demands for funds. As long as more of the same is fundable, that's what'll be proposed.

Other approaches, going under names like 'network' or 'systems' biology will likely make progress, and probably can stimulate the opening of the conceptual door, since molecular interactions are what life is all about (as we argue as a major theme of The Mermaid's Tale). However, the network studies to date mainly show just what GWAS finds and evolutionary biology predicts: things can be genetic in the sense that many interacting genes contribute, but are not easily targetable in terms of 'personalized' medicine related to individual genotypes.

So, if there's something clearly genetic in your family, see a legitimate genetic counselor. Otherwise, put away your wallets: Tea leaves and crystal balls are almost as effective as DNA sequence for predicting risk, and much, much cheaper. if you're in a truly high-risk family, see a real genetic counselor. Meanwhile, to stay healthy, eat right, exercise, and stop worrying: Drink your tea, but toss away the leaves!


Jennifer said...

the opening pun was totally uncalled for!

Ken Weiss said...

Well, the story of genetic services has been a soap opera, and it still is. If anything, there's been a lot of intake at the (money) devouring end, but constipation of results at the other end. Maybe one could grant a kind of goat-like production of a few berries, but not the load that the purgative has promised.