Sun et al. sequenced a cohort of 584 schizophrenia patients and 549 controls to look for rare variants in the NLGN2 gene. They found several missense variants, but no significant excess in cases (8/584 vs. 2/549, p = 0.07). Polyphen predicted one (a case-only variant) to be probably damaging, and another to be possibly damaging, and the rest to be benign. Their own extensive functional work showed the “possibly damaging” one to have no obvious effect on function; rather, it was a loss-of-function mutation. The patient had no family history of mental illness, yet the mutation was shared by the one relative from whom they could obtain DNA—an unaffected brother—suggesting that it was inherited from one of the unaffected parents.

Overall, although the variant certainly seems to be functional, there isn’t really any good evidence that it is involved in schizophrenia in this patient. The gene is a good candidate, but there was no evidence for an excess of mutations in the schizophrenia patients versus the controls. The family history, the presence in the unaffected brother, and the fact that the affected brother apparently had no “environmental hits” that were not shared by other members of his family suggest that it has no effect on schizophrenia susceptibility. Other models in which the gene variant acts only in concert with one or more other variants (in other genes) not seen in other family members are certainly possible. However, the weight of evidence as presented is not convincing.

With the advent of new technologies such as Illumina’s MiSeq, as well as genomewide next-generation sequencing, studies like this will become more and more common, and it can be difficult to know what is relevant to report. Obviously, we don’t want to miss anything of importance, and it is clear that it will take the combined efforts of all of us to elevate rare associations to statistical significance. On the other hand, we do not want a situation where the literature is once again packed with false-positive candidate gene reports. It seems like a good solution would be to set up some minimum standards that a report must meet before publication is warranted. These could include significant excess of rare variants in cases versus controls, or some proven association with psychosis, either through a mouse model, familial segregation, or presence in multiple cases (with some basic stratification check, perhaps). Other suggestive findings could be reported in a central Web-based repository which would serve as a “storehouse” for evidence until it reaches a critical mass and could potentially be reported as a collaborative finding.

View all comments by Anna Need