This is the first report of exome sequencing of sporadic cases of schizophrenia and for that reason is interesting. There are likely to be many more of these sorts of papers appearing over the next few months, and it will be informative to compare and tabulate the results once they are available.
Of most interest to me was that none of the reported de novo mutations was replicated among any of the individuals sequenced, indicating that if these de novo mutations in the exome are pathogenic for schizophrenia, the genetic landscape is vastly more complex than previously imagined.
View all comments by Bryan Roth
Xu et al. (2011) tested the hypothesis that de-novo exon mutations play a major role in schizophrenia by sequencing the exomes of 53 sporadic case trios and 22 unaffected control trios. The experimental procedures for mutation identification were well done technically. However, a number of issues deserve closer consideration.
First, a major study design in human genetics is the evaluation of pedigrees densely affected with a disease under the assumption that etiological variants are more likely to segregate in these so-called multiplex pedigrees. Xu et al. took a very different approach by studying schizophrenia cases with no history of schizophrenia or schizoaffective disorder in their first- or second-degree relatives. Their assumption is that a deterministic exonic mutation occurred that was necessary and sufficient for the development of schizophrenia. In effect, the assumption is that these cases represent different Mendelian forms of schizophrenia. Moreover, as cases are heterozygous for the de-novo mutations, the authors make the fairly strong assumption that the de-novo variants identified act in a dominant mode.
Although not discussed in the paper, there are other models by which schizophrenia can occur sporadically. Indeed, sporadic cases could be more likely to have schizophrenia due to environmental etiologies (e.g., head trauma, CNS infection, obstetric trauma, cannabis use, psychotogenic drug use, etc.). Some of these alternative etiologies can be very difficult to uncover clinically, and it would have been helpful to read more about the authors’ efforts to exclude environmental etiological factors.
As another example, there is now fairly strong evidence that schizophrenia is highly polygenic with a large number of common genetic variants of subtle, probabilistic effect (the Psychiatric GWAS Consortium’s schizophrenia mega-analysis is in press in Nature Genetics). It is possible that some of these sporadic cases resulted from matings between two unaffected parents with moderate to high numbers of schizophrenia risk loci.
Second, in regard to the results, the proportion of de-novo mutations was somewhat higher in cases than in controls, although the difference was not statistically significant. This is not really in line with the main hypothesis, and there could have been more discussion of this result.
The authors then used computational algorithms to predict the proportion of deleterious de-novo mutations, and found a significantly higher proportion in schizophrenia cases than controls. No functional data were presented. Thus, the trustworthiness of this observation is dependent upon the robustness of the algorithms that predict the functional consequences of a mutation. Are these in-silico predictions accurate? These algorithms are definitely imperfect (it is not uncommon for three different algorithms to give different answers). Without additional evidence and functional data, the statement that these deleterious mutations “have a high likelihood of causation with respect to schizophrenia” is assumption-laden and has not been proven.
Third, the author concluded that “de-novo mutations account for more than half of the sporadic cases of schizophrenia.” This claim is unquestionably premature, and not supported by the data.
The single most important lesson learned in human genetics in the past generation is the need for replication. The authors have made an intriguing observation, but, without replication, it is difficult to know if this is secure knowledge or merely another high-profile, false-positive finding resulting from some cryptic bias. Furthermore, the study samples were Afrikaners, a group with an unusual population history, and it is an open question whether the result would generalize to samples elsewhere in the world. Finally, the sample sizes are small and could be subject to any number of sampling biases.
View all comments by Patrick SullivanView all comments by Jin Szatkiewicz