Quantification of Genetic Risk of ASD for a Planned Child Based on the Genomes of Its Parents

Background: A key question that many parents planning to start a family confront is: what is the probability that this child we are planning to conceive will develop ASD? While, many copy number variants (CNVs) and other genetic defects of high penetrance for ASD have been identified in the literature, and numerous genome wide association studies have identified sets of single-nucleotide polymorphisms (SNPs) that confer risk for ASD, no study that the author is aware of has attempted to quantify the overall genetic risk per genome. With a genetic risk per genome quantified, one could then go a step further and examine the question of what is the genetic risk of ASD in a child that has yet to be conceived.

Objectives: Given two genomes representative of potential biological parents, determine the risk that a child conceived from these two individuals will develop ASD.

Methods: A genome wide risk model was constructed to represent genetic risk of ASD given a specific genome. To start with, a review of the literature was conducted to determine a set of variants that confer ASD risk. Variants in this context included CNVs and SNPs. Full chromosomal duplications or deletions were excluded as they generally result from spontaneous mutation. The search was broad in the sense that it was conducted to include autosomal variants, X-chromosome variants, Y-chromosome variants, and mitochondrial variants. For each variant, measures of relative risk and prevalence were obtained from the medical literature or in some cases estimated. In the case of SNPs, the measure of prevalence used was the global mean effect size. A model of overall risk was constructed assuming risk associated with any given variant is independent of the risk of any other variants that do not affect the same segment of DNA. In other words, risks are generally assumed to be multiplicative. Care was taken in this model to avoid double counting risk of specific SNPs with risk of CNVs covering the same section of DNA. This model was calibrated for each sex based on the estimated prevalence of ASD per sex from the literature. Using this genome wide risk model, a separate model of ASD risk in progeny was constructed given two genomes (biological parents) as inputs. This progeny risk model calculates a risk of ASD for a child of each sex using the parents’ genetic data, and data from the literature on the probability of spontaneous mutation resulting in CNVs. The overall ASD risk in the progeny was then computed as a weighted average of the ASD risk of each sex.

Results: The progeny risk model provides a seemingly reasonable quantification of ASD risk. It would be desirable to test it with out of sample data. Its limitations include: no adjustments for non-independent interactions between variants, and no adjustments for environmental or epigenetic risks.

Conclusions: This study provides a practical application of genetic risk to a real-world question that parents confront when they are considering conceiving a child.