| Resumo: | Autism Spectrum Disorders (ASDs) represent a group of childhood neurodevelopmental disorders characterized by three primary areas of impairment: social interaction, communication, and restricted and repetitive patterns of interest or behavior. Although autism is one of the most heritable neuropsychiatric disorders, most of the known genetic risk has been traced to rare variants. Genome-wide association studies (GWAS) have thus far met limited success in the identification of common risk variants, suggesting that ASD may result from the interaction of many variants with low or moderate individual risk, which cannot be detected in current GWAS in a single SNP analysis framework. Recently, molecular interaction networks have been integrated with high-throughput expression data, and the success of this application has been demonstrated through the identification of biologically meaningful subnetwork markers that are more reproducible and with a higher prediction performance. To identify subnetworks implicated in autism and with predictive value for autism diagnosis we have applied a network-based approach to the Autism Genome project consortium GWAS. We have integrated family- based association data from 2588 ASD families genotyped for 1 million single-nucleotide polymorphisms (SNPs) with a Human Protein-Protein interaction (PPI) network. We show, in line with observations in other complex diseases, that the proteins encoded by top genes (genes including one or more SNPs with a Transmission Disequilibrium Test P<0.01 or 0.005) are significantly closer to each other in a PPI network, suggesting that they are functionally related. Furthermore, these proteins were found to preferentially directly interact with each other, and were connected in a significantly larger component than random expectation, indicating that they are involved in a small number of interconnected biological processes. Having validated our initial assumption that autism-associated genes are confined to a limited number of biological processes, we searched for subnetworks that locally maximize the proportion of genes with low P-values in the GWAS dataset. Validation of the results in an independent GWAS and determination of prediction value of these subnetworks are underway. With this approach, we expect to identify biological processes associated with increased susceptibility to ASD, and eventually to derive clinically useful predictive markers
|
|---|