Improving DeepKinZero

Term: 
2018-2019 Summer
Faculty Department of Project Supervisor: 
Faculty of Engineering and Natural Sciences
Number of Students: 
3

Protein kinases play a critical role in the regulation of signaling pathways by catalyzing the phosphorylation of their substrates.  High-throughput experimental methods can identify of phosphorylation sites on proteins at scale;  yet, determining which kinase is responsible for catalyzing the phosphorylation reaction on a given site remains a significant challenge. Existing computational methods rely on knowledge of known sites of a kinase to make therefore their application to understudied kinases is not straightforward.  In our recent work, we cast this problem as a multi-class classification task and developed the first zero-shot learning approach to predict the kinase acting on a phosphosite for kinases with no known phosphosite information.  DeepKinZero learns deep representations of the local sequence surrounding the phosphosite using a bidirectional RNN. Using the kinase characteristics,  transfer knowledge from kinases with many known sites to those kinases with no known sites through zero-shot learning.  This PURE project aims to improve the current performance of the model with additional information on kinases and the phosphosites.

Related Areas of Project: 
Computer Science and Engineering
Molecular Biology, Genetics and Bioengineering
​Mathematics