WCPG 2013—The Future of Psychiatric Genetics: Part 1
This is Part 1 of a two-part series. See Part 2.
November 15, 2013. As the 2013 World Congress of Psychiatric Genetics wound down, a large crowd assembled at the Boston World Trade Center on the morning of Monday, October 21, to hear the third and final plenary panel session of the conference. Conference co-chair Lynn DeLisi cited four exciting days filled with information about new data. We now know that there are well over 100 common alleles that confer risk for schizophrenia (see SRF related conference story) and have heard a lot about new tools to study psychiatric genetics, such as stem cell biology and genome engineering (see SRF related conference story).
“Before we disperse back to our clinics and labs and offices, we thought it was important for us all to get together and really have a discussion,” added panel moderator and conference co-chair Jordan Smoller. He joined a panel of experts to answer critical questions such as, How do we complete this picture? Where should we be placing our investments? How do we ultimately translate these things into new strategies for treatment and risk prediction?
Still a long way to go
In the first half of the panel session, talks focused on how to continue the process of genetic discovery. Patrick Sullivan from the University of North Carolina in Chapel Hill began with what he called the “Ikea” metaphor for psychiatric genetics. He suggested that the field needs three types of information: the parts list, how to put the thing together, and how to use it. The parts list in this case is the genetic architecture—the number, type, and frequency of risk loci; how the loci act and interact (both with each other and the environment); and how multiple disorders may share the same loci.
The first goal, then, is to identify a “good enough” parts list—a goal that is achievable on the three- to 10-year horizon, he said. But uncovering every genetic variant involved in psychiatric illnesses is an exercise in diminishing returns, he added. The objective is not to explain heritability (an imprecise measure), but rather to gain biological insight into diseases in order to develop rational therapies.
Discovering the parts list requires consortia, Sullivan said. The Psychiatric Genomics Consortium (PGC) currently includes full GWAS data from 170,000 subjects. With an eye on NIMH director and fellow panelist Tom Insel, Sullivan suggested that with sufficient funding that sample size could even be doubled within the next year, though he joked that the amount of money required would exceed even Insel’s credit limit. There is not one single parts list, said Sullivan. As part of a “divide and conquer” approach, he named four: one for common variants of small effect; one for large, recurrent CNVs; another for exome sequencing; and a final one for genome sequencing.
Stick with it
Sullivan made several general suggestions to the field. “For once, we actually know what works. I think we should do more of it, not less of it,” he said, cautioning the field against being “flighty” and abandoning something that works for something that’s new and unproven. He also called for rigor, imploring researchers to use robust, replicable genomic standards. Diversity is also important, he said. We should come up with new ways to approach phenotypes and genotypes. He issued a note of caution for the exome: “Just because you have … an interesting loss-of-function variant in some really cool gene … that isn’t causality.” It still requires a different line of genetic evidence to support it, he said.
Sullivan ended by posing a question to the audience: If we get to the point where we have these parts lists, and we start to move into the post-translational and neurobiological aspects of the variants, will this spirit of openness continue? “I certainly hope so,” he said.
The next speaker, Peter Donnelly of Oxford University, UK, echoed many of Sullivan’s sentiments and outlined the goals of genetics as threefold: to provide insight into disease biology, improve risk prediction, and stratify patient populations (refine diagnosis and inform therapeutic choices). All three are important, he said, but at the moment the variants we know simply aren’t predictive enough for us to do a good job on the second two, so “the real opportunity that genetics offers is the chance to learn important new biology.”
There is a lot of evidence that psychiatric disorders involve many, many common variants, Donnelly said, and that they all have rather small effects. GWASs work, he added; it’s just a matter of getting large enough sample sizes. He called the work of the Psychiatric Genetics Consortium (PGC) “exemplary” for its large-scale collaborations that have brought together the large sample sizes that “have really made a difference to discovery.”
With the information already in hand, Donnelly offered two ways of moving forward. Other diseases such as Crohn’s disease have demonstrated that identifying around 100 single nucleotide polymorphisms (SNPs), as we have now for schizophrenia, can highlight pathways of interest or in some cases even identify new ones, he said. In addition to this general approach of getting “high-level” pathway information, every single associated variant is a potential clue to important disease biology. Although figuring out how to go from association to biology is very hard, he said, for the purpose of developing new pharmacological treatments, uncovering the whole functional or biological story is unnecessary. We only need to know the gene involved and the direction of modulation needed, he said. He also noted that there is no direct relationship between a GWAS variant’s effect size and its therapeutic potential.
Turning to rare variants and sequence-based discovery, Donnelly noted that “there aren’t many easy wins.” Rare variants with large effects—the so-called “Goldilocks mutations”—have been very few and far between. While GWASs can test each variant individually, statistical power in sequencing studies depends on allele frequency, so rare variants usually cannot be studied individually. This problem has led to the practice of combining variants within natural units (genes or pathways); however, this approach is also flawed, according to Donnelly, because within a unit some variants may matter and some may not. Nevertheless, focusing on loss-of-function variants makes the process easier and should be prioritized, he said, because they are all likely to have the same effect and thus can be easily combined.
The next wave of studies will involve genotyping in large, prospective cohorts, Donnelly predicted, pointing to the UK Biobank project in which he is involved. It’s an exciting time, he said, though “we’ve learned that it’s much harder than we thought it might be.”
To conclude the first half of the plenary session, panelists took a few comments and questions from the audience. David Curtis (following Sullivan’s advice to distrust others’ opinions) said that he “fundamentally, really strongly disagreed” with both Sullivan and Donnelly. Given that the 100+ hits already identified have such tiny effects, what is the point of ever bigger sample sizes? he wondered. With all the talk of genetics, Naomi Wray and Robin Murray pointed out the need to look at the influence of environmental factors. Sullivan agreed: “It’s never been the case that it's genes or environment; it’s always been ‘and’.” Going forward, large longitudinal studies will be important in ferreting out the impact of the environment, he said.—Allison A. Curley.