Systems Biology Approaches to Determine Genetic Risk Factors for Autism Spectrum Disorders
Objectives: Using systems biology approaches, we aimed at identifying genetic patterns and networks that shape the autism phenotype as defined through the Autism Diagnostic Instrument Revised (ADI-R). Our approach includes identification of underlying genes influenced by associated common and rare variants integrating available knowledge on eQTL, gene-expression during brain development, regulatory sequence patterns as well as protein-protein network information. Finally, we model the phenotypic presentation using the network state of an individual patient.
Methods: Regression modeling predicting ADI-R derived factors was performed using the Autism Genome Project (AGP) data set with whole genome single nucleotide polymorphism (SNP) data. Modeling was further supported by machine learning approaches. Underlying network model construction was based on brain expression data (Kang et al 2011), eQTL and mQTL data (GeneVar, Gibbs et al 2010) and protein-protein interaction database (STRING, GeneMania). Models were validated and further refined by integrating information on functional rare variants in an independent German ASD data set with whole genome SNP data, CNV and exome sequencing data (Autism Sequencing Consortium). Both dataset are matched for gender and ethnicity, and all models were controlled for ethnicity.
Results: Here we show preliminary results: We reproduced previously published ADI-R factor analysis (Liu et al 2011) in both the AGP and the German dataset supporting a six factor (Eigenphenotypes) solution corresponding to Joint attention, Social interaction and communication, Non-verbal communication, Repetitive sensory-motor behavior, Peer interaction as well as Compulsion/Restricted interests. A total of 730525 SNPs were included in regression modeling predicting the extracted six factors. In addition, 6 components controlling for ethnicity based on the genetic background were included. Results of Eigenphenotype modeling and subsequent network analysis are presented.
Conclusions: As ASD is a complex neuropsychiatric disorder with a varying genetic heterogeneity, our research will aid in discovering the association of SNPs with disease phenotype. This project will contribute to a deeper insight into ASD pathomechanisms and potentially provide predictive models for clinical setups.