Prediction of Genes Associated With Autism Spectrum Disorder Using Sequence And Graph Embedding Methods

Inácio, João; Vilela, Joana; Marques, Ana Rita; Santos, João Xavier; Rasga, Célia; Vicente, Astrid; Martiniano, Hugo

http://hdl.handle.net/10400.18/11126

Utilize este identificador para referenciar este registo.

Nome:	Descrição:	Tamanho:	Formato:
HugoMartiniano_Posterint.pdf		1.59 MB	Adobe PDF	Ver/Abrir

Contacte-nos

Autores

Resumo(s)

Introduction: Autism Spectrum Disorder (ASD) is a complex neurodevelopmental disorder, characterized by a strong genetic component, heterogeneous phenotypic presentation and complex genetic architecture. [1, 2] Identification of ASD risk genes and biological pathways is key to understand this disorder.[3, 4] Despite the large number of genes associated with ASD, its genetic etiology is still not fully understood. To address this challenge, we developed a machine-learning approach for ASD risk gene prediction. Methods: We developed a Machine Learning–based approach to identify and rank ASD-associated genes. Results and discussion: Prediction of ASD-associated Genes The best model was used to create a ranked list of ASD-associated genes. Using the top decile of the predicted genes, we performed a decile enrichment analysis using Phenotypes from the Human Phenotype Ontology (HPO) database and gene lists for neurodevelopmental disorders and psychiatric disorders unrelated to ASD. Conclusions: We developed a machine learning approach based on sequence and Graph embedding Methods to predict ASD-associated genes. Our approach compares favorably with state-of-the-art approaches in identifying genes targeted by loss-of-function (LOF) mutations in the MSSNG and Simons Simplex Collection cohorts and shows specificity towards ASD-related phenotypes.