Skip to content

MSc Thesis Defense: Md. Mohaiminul Islam

When:
Monday, 4th December 2017 2:00 pm
Where: 315 Buller Bldg

Title: Deep Learning Models for Predicting Phenotypic Traits from Omics Data

Abstract:

Computational and statistical analysis of high throughput omics data, such as gene expressions, copy number alterations (CNAs) and DNA methylation (DNAm) has become very popular in cancer studies in recent decades because such analysis can be very helpful to predict whether a patient has certain disease or its subtypes. However, due to the high-dimensional nature of the data sets with hundreds of thousands of variables and very small numbers of samples, traditional machine learning approaches, such as Support Vector Machines (SVMs) and Random Forests (RFs), have limitations to analyze these data efficiently. In this thesis, we propose multiple deep neural network (DNN) models for predicting molecular subtypes of breast cancer patients, which include the status of estrogen-receptor (ER): ER+ and ER- and the status of PAM50 subtypes: luminal A, luminal B, HER-2 enriched and basal-like. In addition, we use epigenome-wide DNAm profiles of before and after medication interventions (called pretreatment and posttreatment, respectively) to predict triglyceride concentrations for peripheral blood draws at visit 2 (using pretreatment data) and at visit 4 (using both pretreatment and posttreatment data). We demonstrate that our proposed DNN models are superior to SVM, RF, and deep belief network (DBN) in terms of prediction performance. Our experimental results show that integration of multi-omics profiles into DNN-based learning methods can improve the prediction of the molecular subtypes of breast cancer. This study also suggests that the DNN approach has advantages over other traditional machine-learning methods to model high- dimensional epigenome-wide DNAm data and other genomic data. 

Complete Seminar List
© 2011 University of Manitoba Department of Computer Science
Back to top