Site Search
Computer Science


Windsor WaterfrontWindsor Waterfront Park
Windsor Waterfront Park
Xiaobu Yuan, Ph.D.Dr. Xiaobu Yuan
Dr. Xiaobu Yuan
Dr. Scott GoodwinDr. Scott Goodwin
Dr. Scott Goodwin
Dr. Robert KentDr. Robert Kent
Dr. Robert Kent
Lambton TowerLambton Tower
Lambton Tower
Dr. Ziad Kobti lecturingDr. Ziad Kobti
Dr. Ziad Kobti
Christie Ezeife, Ph.D.Dr. Christie Ezeife
Dr. Christie Ezeife
Alioune Ngom, Ph.D.Dr. Alioune Ngom
Dr. Alioune Ngom
Dr. Luis RuedaDr. Luis Rueda
Dr. Luis Rueda
Robin Gras, Ph.D.Dr. Robin Gras
Dr. Robin Gras
Arunita Jaekel, Ph.D.Dr. Arunita Jaekel
Dr. Arunita Jaekel
Imran Ahmad, Ph.D.Dr. Imran Ahmad
Dr. Imran Ahmad
Jessica Chen, Ph.D.Dr. Jessica Chen
Dr. Jessica Chen

Integrative Machine Learning Models for Inferring Pairwise Associations from Biological Databases

Add this event into your calendar using the iCAL format
  • Tue, 07/17/2018 - 1:30pm - 3:30pm

Integrative Machine Learning Models for Inferring Pairwise Associations from Biological Databases

PhD. Thesis Proposal by:

Hetal Rajpura

Date:   Tuesday, July 17, 2018
Time: 1:30 pm - 3:30 pm
Location: 3105, Lambton Tower

Abstract: Machine Learning methods have been widely applied to image, video, speech datasets and have started receiving much attention for mining publicly available biological databases. In the domain of biological association prediction, there exists issues that need considerable attention such as 1) Missing truly reliable negative samples 2) Interactions are perceived as binary on/off relationships 3) Quantitative and Geometrical features of molecules are not considered in building predictive models. We start from a simpler model implementation with - one class support vector machine for learning from labelled and unlabelled data and graduate to evaluating auto-encoders and kernel based regularised least squares methods. Similarity kernels are built from structural, evolutionary and physicochemical properties of drugs and targets and further applied as features in Regularised Least Squares models. Auto-Encoder have been known to learn accurate representations from raw data. It seems intuitive to implement them to learn low-level feature representations from high-dimensional heterogeneous biological data. Thereby, we evaluate the use of Auto-Encoders in representing the biological entities in modelling the pairwise input prediction problem. Different experimental settings are constructed for carrying out cross validations for predicting the unknown pairs. Further, these approaches can be extended to any biological databases such as drug-drug interaction, drug-gene interaction and gene-disease data; however, we will mostly focus on drug-target interaction data.

Thesis Committee:     
Internal Reader: Dr Luis Rueda and Dr Mehdi Kargar         
External Reader: Dr. Siyaram Pandey
Advisor: Dr Alioune Ngom               


See More: