30 June, 2017

Sharing latest application of deep learning in bioinformatics at University of Missouri

The Big Data Institute and Division of Life Science has jointly hosted a seminar on Application of Deep Learning in Bioinformatics by Prof Dong Xu on 30 June 2017. Introduced by Prof Jiguang Wang, Prof Xu from Department of Electrical Engineering and Computer Science, University of Missouri, Columbia presented some of their successful applications on deep learning to the HKUST community.

Deep learning, as the cutting-edge technology in machine learning, has produced remarkable improvements on various computational problems and presents a new opportunity in bioinformatics. The growing amount of various biomedical data also allows deep learning to generate robust models. In this talk, Prof Xu presented several of our successful deep-learning applications:

(1) Developed MusiteDeep, the first deep-learning framework for predicting general and kinase-specific phosphorylation site. MusiteDeep takes raw sequence data as input and uses CNN with a novel two-dimensional attention mechanism;

(2) Developed a deep neural network approach for large-scale prediction of plant mitochondrial proteins by using various types of features, including amino acid composition, protein sequence profile, and gene co-expression information;

(3) Developed a novel feature extraction framework called constrained high dispersal neural networks (CHDNet) for tongue image analysis and health assessment; and

(4) Applied deep learning in analyzing news text for predicting disease trend.