I recommend the Primary Papers

The paper by Allen et al. is a tremendously useful addition to the fields of schizophrenia research, psychiatric genetics, and medical genetics. By efficiently summarizing a tremendous amount of work, Allen et al. have endeavored to provide a "state-of-the-art" summary that most of us, as individuals, struggle to accomplish; they have largely succeeded in their attempt. This manuscript, and the continual availability of the SZGene database, should long serve as invaluable resources for the increasingly complex task of building polygenic models of risk for schizophrenia. Furthermore, these methods, which were initially implemented in the AlzGene database, have clearly generalized quite successfully to SZGene and thus, should be easy enough to scale up to cover many other psychiatric disorders as well. In this way, the contribution to psychiatric genetics, and possibly other disorders outside of psychiatry, is crystalline.

Aside from the database, the contribution of the recent manuscript to the field of schizophrenia research is also tremendous. As pointed out by the authors, several of the significantly associated genes identified by their meta-analyses were never before studied in this manner, so a whole new set of top candidate genes was identified. This work also served to confirm the results of prior meta-analyses from my group and others, which is always reassuring. Application of the HuGENet criteria to grading the detected associations is useful as a heuristic, but it must be kept in mind that that while these criteria reflect a consensus, they also reflect a moving target. One difficulty in implementing grades (especially the "overall" grade) is analogous to difficulties often encountered in meta-analyses when rating the quality of studies, and that is the ambiguity of ratings. Thus, on a seven-point quality scale (or a three-letter-grade scale), a score can be arrived at by a variety of combinations of flaws or strengths, but similar scores may not (often do not) reflect identical strengths and weaknesses of the graded studies. For example, I, for one, am not certain that having a relatively low number of minor alleles reflected in a meta-analytic result (especially if it is a rare variant) is as big a decrement as the pooled OR dropping from significance when the initial study is omitted.

Nevertheless, I reiterate that the use of this heuristic grading system is helpful, but should be taken with a grain of salt. Overall, the paper and its conclusions are a great contribution to this field and warrant mass attention. The ultimate question, not yet addressed here but apparently on the horizon, is how well the emerging GWASs detect these "positive control" associations, or we might say how well these hypothesis-driven results stack up against new candidates to emerge from the high-throughput generation of novel hypotheses....

View all comments by Stephen J. Glatt