The paper in Nature last week by the Psychiatric Genomics Consortium's Schizophrenia group astonished many with both its findings and its huge amounts of media coverage. This has made many psychiatric genetics researchers very happy, bringing a not-inconsiderable amount of relief that they have been proven at least partially correct.
For many years there was a considerable lack of success in schizophrenia genetics research. It was a controversial field, with many seemingly important findings not being replicated in later studies. In the early 2000s, the field reached its nadir and it became obvious that a different approach was needed. This was found in genome-wide association studies (GWAS), where DNA microarray chips, printed with millions of short DNA fragments, were used to assay or 'genotype' a dense map of information. But success was not instant. From the first GWAS studies in schizophrenia it soon became apparent there were no common variants that were a large influence on risk, but rather that there were thousands of variants of very small effect that together acted to increase or reduce risk. Some critics, of course, regarded this as an excuse and recommended that no further GWASs should be funded.
The first signs of success started to emerge in 2008 with a paper from Michael O'Donovan and colleagues. In 2009, the SGENE consortium and the International Schizophrenia Consortium published papers on the other variants associated with the illness. These groups then combined their efforts and began publishing under the Psychiatric Genomics Consortium banner with papers in 2011 detailing five variants and 2012 with 18 variants. However, real progress began when the sample size exceeded 15,000 cases or so, with an inflection point being reached where, instead of discovering one variant every couple of thousand cases, variants began to be discovered when, on average, another 250 cases and 250 controls were added to the sample size. What happened? Well, I explain it as 'the power met the real effect size' – once the size of the sample has increased to this point, the power to detect variants reaches a point where it can detect the effects of the modest size at which lots of schizophrenia-associated variants are operating.
Thus, eventually the field got to the point where they published last week's paper on the finding of over 108 regions associated with risk of schizophrenia. The key to this was unprecedented sample size – they published on over 37,000 cases and 113,000 controls. More to the point, the study has implicated several new mechanisms in the aetiology of the disorder. Many believe it will start a new revolution in treatment, with enough hard and diligent work. Schizophrenia treatments have not really changed since the widespread adoption of atypical antipsychotics in the 1970s, which themselves represented only an incremental improvement on the older typical antipsychotics identified in the 1950s. Now, novel mechanisms have been identified, such as calcium channels and B-cell immune function, where there is already a huge store of knowledge about the biochemistry, drugability and function of the targets identified via our colleagues in the hypertension and immunology research community.
Drug development in psychiatry is notoriously hard, however, with many pharma companies having flown from the field. At the very least, it may be five to ten years before any new treatments will emerge. However, with so many new potential drug targets to choose from, the hope is that - although many leads will fall to the wayside due to toxicity or other problems - at least a handful will be successful. If this comes to pass, these drugs will, in all probability, have a different biochemical mode of action compared to current antipsychotic drugs. This is something desperately needed by the field as schizophrenia is currently only partially treated, and some patients remain treatment-resistant or develop complications such as persistent cognitive impairment on which the effect of antipsychotics is minimal.
The other, and perhaps more controversial, aspect to address is whether or not these findings have any predictive power. Here, critics can relax as the genetics of schizophrenia are not, and probably will never be, developed enough to screen the general population. This is primarily because of something called the base-rate fallacy whereby a partially sensitive and accurate test will result in many more false positives than true positives when the condition being screened for is rare (schizophrenia affects around 0.4-0.8 percent of people). Thus population screening would be not only bring with it philosophical questions but would be also utterly useless.
However, where genetic prediction may have clinical utility is in identifying people in situations where they have a higher likelihood of developing the illness. One example would be on someone's first admission to a psychiatric hospital, when a test giving the risk profile of someone for schizophrenia may allow psychiatrists to prescribe the correct medication earlier; something that has been shown to result in a much better prognosis. The potential benefit of between being able to distinguish between different psychiatric outcomes early on in treatment is large and will be investigated by many groups over the next few years. The approach could be also used for selecting patients very early in the course of illness for clinical trials of newly developed drugs.
The true measure of this study's success will be whether we are talking about new and more effective clinical treatments in five to ten years as a result of it. To do that, much of the subsequent functional and drug work will need funding and pharmaceutical companies will need to reinvest in the field. However, that note of caution aside, we now know so much more about schizophrenia's causes than we did a year ago that this paper should be celebrated. Given that similar efforts are underway to increased sample size across a host of psychiatric disorders, the future looks a lot brighter than it once did.