System medicine approach to improve diagnosis and prognosis in Autism Spectrum Disorders (ASD), based on extensive genomic, biochemical and clinical data

BioSys-PhD, Biological Systems: Functional and Integrative GenomicsAutism spectrum disorder (ASD) is a neurodevelopmental disorder of well known complexity. ASD is characterized by impaired social interaction and communication and by stereotyped behaviors, and a high heterogeneity in clinical and genetic presentation. It is hypothesized that such complex heterogeneous phenotypic behaviors are associated with genetic factors. To dissect the complex correlations between phenotype and genotype in ASD, in the current study we will use powerful machine learning and data mining algorithms, like decision trees. We will integrate clinical information (from the diagnostic instruments ADI-R: Autism Diagnostic Interview-Revised and ADOS: Autism Diagnostic Observation Schedule, as well as adaptive behavior scale VABS: Vineland Adaptive Behavior Scale and cognitive scales adapted to age and cognitive level) and genetic data (Copy Number Variants, CNVs) of 3000 ASD individuals with ASD. Data on this patient cohort was obtained by the Autism Genome Project international consortium, which included 335 Portuguese patients from our dataset. This analysis will identify autism behavior associations with genetic risk factors, and eventually allow categorization of patients and prognosis according to genotype. We will initially assess the effect of deletion and duplication events and de novo and transmitted CNVs in disease clinical presentation, and progress to analyze the association of CNVs containing candidate genes for ASD with disease phenotype. So far, the etiology of autism is not well understood due to interactions between multiple factors. Genetic, metabolic, gastrointestinal, immunological and neurobiological factors have been associated with ASD etiology. Therefore, we will use a system biology based approach for ASD analysis, which will integrate genetic, miRNA, neurobiology and clinical data to determine how multiple factors can influence the autism heterogeneity. This work will improve the accuracy of data mining techniques, by building specialized classifiers based on a machine learning approach), and by applying semantic enrichment analysis. These classifiers will help in rapid diagnosis of ASD. Moreover, we will provide a framework for autism analysis with knowledge graph based data organization. This framework will enlist classifiers, feature selection and cross validation methods for ASD analysis. We will also provide a comparative and testing phase to cross check the accuracy of framework. ASD is a complex disorder, therefore enhanced understanding of associations at multiple levels (genetic, miRNA, neurobiology, clinical and behavioral), will be useful to assist in ASD diagnosis and prognosis.Muhammad Asif, Doctoral Research Fellow for Fundação para a Ciência e Tecnologia do Ministério da Ciência, Tecnologia e Ensino Superior SFRH/BD/52485/201