Wen, Alshikho and Herbert have provided an interesting network analysis of candidate ASD genes as curated by the SFARI Gene database. The pathways they identify—particularly calcium and Ras/MAPK signaling—are plausible and fit reasonably well with previously published analyses. That said, as one of the people involved in the curation and production of the SFARI Gene database, I want to add a note of caution. The authors have taken an ‘unbiased’ approach by including in their analysis every gene listed in the database. While I am generally in favor of unbiased approaches, in this instance there is a chance that it could generate misleading results. Most (though not yet all) of the genes in the human gene module of SFARI Gene have been assigned a score. This score suggests the likelihood that the gene is actually a risk-conferring gene for ASD based on a careful assessment of the degree of statistical support for the gene’s association with the disorder. Contrary to the suggestion of the authors, this score does not reflect our notion of what really constitutes a reliable diagnosis of ASD, as our scheme assumes all reported diagnoses in the literature are reliable. Rather, our approach is based on what we do in fact consider to be a gold standard for relevance in human genetics—namely a careful comparison of the frequency of mutations (or polymorphisms) in a particular gene in cases versus controls. While some of the details of this approach might differ depending on who is doing the assessment, we think this general framework is widely accepted. The authors correctly point out that not all of the genes on the SFARI Gene list are accompanied by a score, and this is something we are working to rectify. I would like to point out, however, that our scoring efforts prioritize genes that clearly have more support than others, and as such the genes in the top three categories are essentially complete. We suggest that researchers doing similar analyses focus on these genes (most of which are reported in cases of idiopathic autism), since they are much more likely to contain bona fide risk genes than those in the bottom three categories. Finally, I would note that we are currently re-designing the SFARI Gene site, to make it easier to understand and use the scoring system that has been implemented. This new version of the site should be launched by June of this year.

Alan Packer
Simons Foundation Autism Research Initiative