The dream -- or nightmare? -- of the < $1000 genome

post by Bill Gardner

There are great hopes that the rapidly falling costs of gene sequencing will lead to a revolution in medical care.

“For all of human history, humans have not had the readout of the software that makes them alive,” said Larry Smarr, director of the California Institute of Telecommunications and Information Technology, a research center that is jointly operated by the University of California, San Diego, and the University of California, Irvine, who is a member of the Complete Genomics scientific advisory board. “Once you make the transition from a data poor to data rich environment, everything changes.”

The idea is that with truly inexpensive gene sequencing technology, we could conduct a lot of epidemiological studies that would greatly enhance our ability to identify persons at risk of disease, so that gene sequencing becomes the foundational technology for truly personalized medicine.  Each of us can afford to have the data on our own genome to guide our health care decision making. As a bi-product, our greatly increased ability to measure disease risk would greatly change, if not destroy, the health insurance system.

These things may eventually happen, but do not hold your breath. First, it turns out that information from gene sequences that strongly predicts human disease is less common than many imagine. Looking at the genetics of height is informative. Height is highly heritable and can be measured very accurately. So one would expect gene sequencing would enable us to find the key determinants of height. But extensive research with tens of thousands of persons and scanning more than half a million SNPs led to the identification of a few tens of factors that were replicable across studies, each of which contributed (on average) about 0.5 cm to height. If this result is typical, it suggests that your gene sequence will provide only limited information about your health. 

Second, the cost of gene sequencing may be falling rapidly, but the other costs of doing genetic epidemiology are not. You also have to follow patients over a considerable period of time, and measure their health outcomes. Hence genetic epidemiological studies will continue to cost millions of dollars, and be based on limited samples of data. Nor can you come to a conclusion about a health risk based on just a single study -- you have to replicate the finding in one or more additional studies. Therefore, very cheap sequencing may not produce as much genetic epidemiology data as some predict.

Instead, we may be in a situation where we have millions of elements of a gene sequence that may be related to a disease, but only a few that actually are, and those latter have only small effects on whether you have the disease. Moreover, we will have only tens of thousands of persons about whom we know the disease outcome for a given disorder. The combination of many potential risk factors, small effect sizes for these factors, and few cases relative to the number of potential risk factors leads to a large chance that an association between a genetic factor and a disease reported in a single study will prove to be false.

This is a chronic problem in epidemiology. What's new, however, is that there will be many people who have sequenced their genes and own that data. So imagine you learn that a Dr. Jones holds a press conference to announce that she has found that factor X is associated with Fulminant Necrotic Something or Other (FNSOO). You consult their genome and discover that you have factor X. What you have gotten for your trouble is the fear that you could get FNSNOO. Maybe Dr. Smith will find, next year, that in his data factor X is not associated with FNSNOO. Will you get the update? Failures to replicate are notoriously hard to publish, and Smith's non-finding may never reach the literature.

We should do the research, because there may be important risk genes out there. Moreover, everyone has a right to their data. I'm just saying that until we really understand the fundamental biology of how genes affect disease, the gene sequencing enterprise has the chance of doing more harm than good.

Comments

[Will]

This is a very useful counterpoint to many of the genetic chapters in "The Creative Destruction of Medicine". I was left incredulous by how little value-add the collection of SNPs really provide to a person. I think the optimism is very misplaced, especially as we begin to factor in epigenetics/epistasis.

That said: A slightly higher risk in FNSOO (which already has a low prevalence) isn't a significant health problem. BUT, when you're dealing with a disease with already high prevalence (breast cancer, colon cancer), these increases of 30, 50% become quite significant.

But then again, as this catches on, the data becomes more useful so...

[Bill Gardner]

Thanks. Your point about epigenetics is exactly right. The benefit to the patient will have to be demonstrated; those demonstrations will themselves be very expensive.

Cheap gene sequencing will accelerate fundamental biology, which is awesome. It's hard for me to see that there will be significant near term clinical benefits.