Skip to main content

BCB Grad Seminar Hongzhu Cui

Thursday, November 17, 2016
12:00 pm
Floor/Room #: 
SL 402


Boosting Gene Expression Clustering with System-Wide Biological Information and

Deep Learning


Abstract - Gene expression analysis is a widely used and
powerful method for investigating the transcriptional behavior
of biological systems, classifying cell states in disease, and
elucidating molecular mechanisms. However, most of the
current clustering methods do not take prior biological
information into account. In this project, we propose a novel
protocol for gene expression clustering analysis. The protocol
benefits from the deep learning architecture, which provides
more accurate high-level representation of the feature sets, and
incorporates prior biological information into the clustering
process. We tested our method on two distinct gene expression
datasets and compared the performance of our protocol with
two commonly used clustering methods. The results shows that
our proposed protocol outperforms the classic clustering
methods on both labeled and unlabeled datasets. Our results
demonstrate that the deep learning framework could
generalize specific properties of gene expression profiles and
confirm the hypothesis that prior biological network
knowledge could be helpful for the gene expression clustering