TY - JOUR T1 - Simple genetic models for autism spectrum disorder JF - bioRxiv DO - 10.1101/017301 SP - 017301 AU - Swagatam Mukhopadhyay AU - Michael Wigler AU - Dan Levy Y1 - 2015/01/01 UR - http://biorxiv.org/content/early/2015/03/30/017301.abstract N2 - To explore the interplay between new mutation, transmission, and gender bias in genetic disease requires formal quantitative modeling. Autism spectrum disorders offer an ideal case: they are genetic in origin, complex, and show a gender bias. The high reproductive costs of autism ensure that most strongly associated genetic mutations are short-lived, and indeed the disease exhibits both transmitted and de novo components. There is a large body of both epidemiologic and genomic data that greatly constrain the genetic mechanisms that may contribute to the disorder. We develop a computational framework that assumes classes of additive variants, each member of a class having equal effect. We restrict our initial exploration to single class models, each having three parameters. Only one model matches epidemiological data. It also independently matches the incidence of de novo mutation in simplex families, the gender bias in unaffected siblings in simplex populations, and rates of mutation in target genes. This model makes strong and as yet not fully tested predictions, namely that females are the primary carriers in cases of genetic transmission, and that the incidence of de novo mutation in target genes for families at high risk for autism are not especially elevated. In its simplicity, this model does not account for MZ twin concordance or the distorted gender bias of high functioning children with ASD, and does not accommodate all the known mechanisms contributing to ASD. We point to the next steps in applying the same computational framework to explore more complex models.Author summary For understanding complex genetic diseases one needs both data and molecular/genetic models. In the absence of any model, it is impossible to do more than summarize observations. A good model will be consistent with much or all of the existing data and puts the data in the context of known genetic principles. Ideally the model will make testable predictions. Where the good models fail often shows the directions that require more thought about mechanisms. In this paper we describe a new computational framework that we use to explore a complex genetic disorder with many gene targets, with both de novo and transmitted variants, and with gender bias. The disorder we consider is autism spectrum disorder (ASD), and our framework rules out some previous models that make unsustainable predictions. We identify a formal model that satisfies diverse epidemiologic and genomic observations. This model makes strong and untested predictions and thereby suggests new studies that would resolve outstanding aspects of autism genetics. ER -