SCHOOL OF COMPUTER SCIENCE

The School of Computer Science is pleased to present…

 

 

Date: Friday December 6, 2019

Time:  2:00pm -4:00pm

Location: Essex Hall 122

 

Mo5fs are short, highly conserved sequences in DNA and proteins that can be used in classifica5on or the understanding of their structures and func5ons. Discrimina5ve mo5f discovery is the problem of finding only those mo5fs that can help to discern one set of sequences from another. This can eliminate false-posi5ve findings by providing a nega5ve set of mo5fs, which are known not to perform the desired func5on. Many mo5f discovery algorithms are known to be 5me-consuming due to their exponen5al-5me complexity, making them suitable only for finding short mo5fs in small datasets. This means that they cannot be scaled to the large datasets available today. Recently, a model called DeepBind, based on convolu5onal neural networks was created to find mo5fs that assist in DNA-protein binding. Basset, used a similar model to find regulatory regions in eukaryo5c genomes. In this work, we build on top of DeepBind to show that the model can be generalized to classify proteins into families. We are also able to find mo5fs of varying lengths and perform mul5- class classifica5on.

 

Thesis Committee: 

MSc Thesis Proposal Announcement

 

5113 Lambton Tower, 401 Sunset Ave., Windsor ON, N9B 3P4 (519) 253-3000 Ext. 3716 csgradinfo@uwindsor.ca