Will newborn whole-genome sequencing be valuable by 2019?

In a comment on my previous post, Tera Eerkes is skeptical about the utility of routinely performing whole-genome sequencing on newborns:

I found this comment absolutely fascinating, given the recent
reports on translational analysis, that indicate an actual lack of
clinical utility of KRAS testing and other drug-gene interactions.

I believe these reports are indicative of a trend, not an exception.

I think it is remarkable that anyone feels, even optimistically,
that we're going to need a genome-wide scan for clinically useful
indicators by 2019.

I predict that there will be a relatively small number of genomic
regions, perhaps less than 100, that will be worth testing by 2019.

If we use a whole genome technology, it won't be because we need all
of that data, it will be because it is just so inexpensive to capture
it.

We're still going to be "all dressed up, with no place to go" though.

I don't completely disagree with Tera - for instance, there is certainly some truth to the argument that whole-genome sequencing will initially become routine more due to cost-effectiveness than to utility (as the cost of large-scale sequencing continues to plummet, it will eventually be cheaper to sequence an entire genome than to perform a series of separate genetic tests - this is certainly likely to be the case by 2019).

However, I think the value of whole-genome sequencing will be much higher than Tera anticipates by 2019, for several reasons.

Firstly, there are plenty of very rare, severe disease mutations that would be clinically useful to detect at an early stage. Most developed countries already screen newborns for a relatively small set of disorders that are common enough for screening to be cost-effective, and actionable (in the sense that clinical outcomes can be improved if they are medically managed from an early stage). Cheap whole-genome sequencing would make it cost-effective to expand that list to include essentially any single-gene disorder.

Many diseases that are currently non-actionable will become so in the near future, as medical science advances; and for many parents, finding out about even non-actionable diseases would be worthwhile, as it would allow them to prepare themselves emotionally and logistically for the onset of disease (of course, parents could always elect not to find out about these diseases).

Secondly, pharmacogenomics is finally coming of age. Even just a couple of years ago, pharmacogenetics was a field that had almost completely failed to live up to its over-hyped predictions: there were few well-replicated markers for drug efficacy or toxicity, and even fewer with any clinical value. This is rapidly changing in the era of genome-wide association studies.

Take two drugs as examples: the antiretroviral abacavir, and the anti-cholesterol agents known as statins. I'll be writing more about these drugs soon, but the major message from recent research is this: both drugs have a low frequency of severe side effects; and in both cases the majority of the individuals who suffer these adverse reactions can be identified by typing a single genetic marker.

The application of genome-wide association studies to other drugs is still in its infancy, but it seems likely that in general the genetic architecture of drug toxicity is much, much simpler than that of common diseases like type 2 diabetes; that means that the identification of clinically useful genetic variants will be proportionally easier. As an example, a recent study found that identifying the abacavir sensitivity variant could have been accomplished with a genome-wide association study of just 15 cases and 200 controls - a laughably small number, given that diabetes studies with tens of thousands of samples are still struggling to find the majority of the risk variants.

Thirdly, it's likely that the next wave of association studies will start to dig up rare, moderate-effect risk variants for common diseases. The current crop of studies are seriously under-powered to detect low-frequency variants, but there is good reason to believe that the lower reaches of the frequency spectrum are enriched for variants with a much larger individual effect on disease risk (odds ratios much larger than 2) than the common variants discovered to date (I'll explain why in upcoming posts). New studies, using either new sequencing technologies or new chips targeting low-frequency variants in very large cohorts, will pick these elusive creatures up.

Rare, moderate-effect variants won't necessarily explain much of the "missing heritability" of common diseases on a population level, but they will be far more useful for predicting individual disease risk. If you're one of the few unfortunate souls who happens to carry two or three of these variants, your risk of disease will be substantially increased - often enough to warrant increased surveillance or lifestyle interventions.

Finally, and perhaps most importantly, routine whole-genome sequencing would provide an incredible resource for further research into genetic effects on health. The current crop of genome-wide association studies have captured most of the easy pickings; finding things like rare variants, or common variants with very small effects, or interactions between genes or between genes and environmental factors, will require studies with enormous numbers of participants - on the order of hundreds of thousands of individuals.

Recruiting hundreds of thousands of people into a special study is possible (and there are indeed large academic consortiums currently doing precisely this), but it's also fiendishly expensive. If governments have the foresight to establish solid infrastructure for linking whole-genome sequence data with electronic medical records and other data sources, with appropriate privacy regulations, then they will have a powerful resource for research and monitoring public health.

In such a system the importance of performing whole-genome sequencing - rather than just a series of isolated genetic tests - is that it allows the generation of truly unexpected associations between genetic variants and diseases. That's a much more effective way of uncovering the complete genetic architecture of human disease than the current piecemeal approach.

As I said in my previous post, these advances are contingent on both researchers and regulators doing the right thing: researchers will need to do a better job of converting their results into clinically relevant tools, and regulators will need to build a thorough legal framework that protects the public from abuse of genetic information, without stifling innovation. If everyone does their job, routine whole-genome sequencing will be valuable both to individual health and to society as a whole.

Subscribe to Genetic Future.