Department of Computer Engineering
CS 590 SEMINAR
Network-based Analysis of Individual Mutation Profiles for Autism Spectrum Disorder (ASD)
Computer Engineering Department
Autism Spectrum Disorder (ASD) is common neurodevelopment disorder affecting 1 in every 68 children. Since early intervention can improve the patients conditions significantly, early diagnosis of ASD is an important task. However, majority of ASD symptoms are related with social interactions and these symptoms are hard to notice at early ages. In order to tackle this problem, instead of just looking symptoms for diagnosis, our goal is to use genomic profiles of patients for early prediction of ASD. This aim comes with several challenges. First, mutations that are informative for ASD prediction are very rare, which yields a very sparse feature vector. Second, unlike a simple classification task, a false positive prediction might have devastating effects on the child and the family. In this project, we use a network smoothing based approach to amplify the signal of informative mutations by propagating the signal over a gene co-expression network. Then, this information is used to predict if a child has ASD or not, while minimizing the false discovery rate (FDR) and maximizing the true positive rate (TPR).
DATE: 10 April, 2017, Monday @ 15:40