A news focus piece in last week's Science about cancer geneticist Bert Vogelstein is right up our alley. The piece begins, "Their lab helped reveal how faulty genes cause cancer, but Bert Vogelstein and [laboratory co-director] Kenneth Kinzler sometimes irk colleagues with their “reality check” comments on genomic medicine." Their point? Whole genome sequencing is not going to be useful for predicting who will and who won't get cancer. And they back this up with a study of disease risk in identical twins, described in an April Science Translational Medicine paper ("The predictive capacity of personal genome sequencing").
Vogelstein has long been interested in characterizing genes that are mutated in tumors, long ago identifying genes associated with the development of colorectal cancer. He and Kinzler when the latter was a student in Vogelstein's lab, showed how the slow accumulation of mutations in previously identified genes, including tumor repressor genes that no longer do their job when mutated, lead to tumor growth.
In the last decade, Vogelstein and Kinzler were the first lab to publish an extensive tumor exome sequence, all the coding regions of breast and colorectal cancers. They identified both known and novel genes involved in tumorigenesis. The work was done in the days before high-throughput sequencing was commonplace, however, and their work was criticized as not having been thorough enough or well-analyzed statistically. Though, their results were subsequently confirmed by others.
Whole genome tumor sequencing is much easier and more complete now, but Vogelstein and Kinzler don't see much more to be gained with it, and they've moved on. Their recent work has involved looking at identical twins to determine whether what they call the "genometype" would allow prediction of disease risk. That is, based on the assumption that monozygotic twins share essentially the same genotype, is it possible to predict risk of disease to a second twin if the first one has it? This of course depends on the extent to which the disease is genetically determined.
This basic observation, that monozygotic twins of a pair are not always afflicted by the same maladies, combined with extensive epidemiologic studies of twins and statistical modeling, allows us to estimate upper- and lower- bounds of the predictive value of whole-genome sequencing.
On the negative side, our results show that the majority of tested individuals would receive negative tests for most diseases. Moreover, the predictive value of these negative tests would generally be small, as the total risk for acquiring the disease in an individual testing negative would be similar to that of the general population.The authors go on to point out that this is consistent with what has been found with GWAS -- many genes explain little risk.
The story is different, they point out, for rare monogenic diseases, where whole genome sequencing has already been shown to be informative -- but then, so have association studies and the like.Thus, our results suggest that genetic testing, at its best, will not be the dominant determinant of patient care and will not be a substitute for preventative medicine strategies incorporating routine checkups and risk management based on the history, physical status and life style of the patient.
Why the cold water is warranted
Most such somatic mutations are never seen clinically. Whether they help or harm, they're just in a single cell, and their effects are swamped by the sea of surrounding cells in the same tissue, that basically have your constitutive genotype at genes relevant to that tissue. If the constitutive genotype confers risk, then basically all cells in that tissue are at risk.
The difference with regard to cancer is that when a bad combination of mutations occurs in a single cell, it doesn't just die or stagger along doing no harm to you, but it proliferates, amplifying the signal of that mutation. It takes many different mutations to transform a cell from normal to cancerous. This is why cancer risk is poorly predictive from your constitutive genotype: most of the changes that lead to disease occur somatically in this or that cell until a bad combo arises in one of billions of cells.
So you'd think at least looking at the tumor cells would show what mutations were important. To some extent that's true, and though it isn't much use in predicting cancer (since the mtuations are found after you already have cancer!), this may provide ideas on how to target the cancer cells. The problem with even that is that a cancer in a single person is continually evolving, rapidly accumulating even more mutations, so that not all cells in the same tumor are cancerous for the same reasons.
You'd have to sample many different parts of the tumor to identify the different variants. And some recent studies have done just that, and shown that different secondary tumors--descendants of the primary tumor, within one patient's body--are genetically different. In part, at least, this is what enables cancer to metastasize, to colonize different parts of the body from the tissue they started in.
But if cancer is therefore not well predicted from your constitutive genome, one might expect that diabetes and heart disease would be predictable because they aren't the same kind of proliferating disorder. But despite what the genome-selling companies would like you to believe, that is turning out not to be true, either, and we have discussed this countless times before, in the context of GWAS and other studies.
Evolutionary implicationsThis is all consistent with evolution as well. The same genomic complexity that makes your traits, but makes finding single genes 'for' the trait difficult, is exactly what means that natural selection is not working very closely on one or a small number of specific genes. If GWAS can't find causal genes for a trait, even if you have the trait, natural selection can't do that either.
This means that traits can evolve adaptively via natural selection in the way Darwin explained, without this being very tractably understood at specific single genes, and indeed the indirect genomic effects of selection that is merely screening traits is what led causation to be so complex in the first place.
There are many parallels between what happens among cells in your body, and individuals in a species, and Ken wrote about that in 2005 in Trends In Genetics, where he discussed ways in which diseases other than cancer might be caused by somatic mutations whose effect could somehow be amplified so you would notice it at the organism level. Diseases like epilepsy were examples discussed there.
Causation may be genetic in the trait or evolutionary sense, but the specific genotypes that are responsible may be difficult or impossible to detect, or so variable among cases and individuals that by and large it's not worth taking that approach--something roughly consistent with what Vogelstein was saying in the story about him.