Proceeding talk – Theme: Genome.
Abstract
Histone modifications are among the most important factors that control gene regulation. Computational methods that predict gene expression from histone modification signals are desirable for understanding their combinatorial effects in gene regulation. This can help in developing “epigenetic drugs” for diseases like cancer. This paper develops a unified discriminative framework using a deep convolutional neural network to classify gene expression using histone modifications as input. Our system, DeepChrome, allows automatic extraction of complex interactions among important features. To visualize the combinatorial interactions among histone modifications, we propose a novel optimization-based technique that generates feature pattern maps from the learnt deep model. This provides an intuitive description of underlying epigenetic mechanisms that regulate genes. DeepChrome outperforms state-of-the-art models for gene expression classification task on 56 cell-types from REMC database. The output of our visualization technique validates the previous observations and allows novel insights about combinatorial interactions among histone modification marks.
Authors
Ritambhara Singh, University of Virginia, United States
Jack Lanchantin, University of Virginia, United States
Gabriel Robins, University of Virginia, United States
Yanjun Qi, University of Virginia, United States