CS Colloquium: Jianzhu Ma, Project Scientist, Dept. of Medicine, UC San Diego
The School of Informatics, Computing, and Engineering (SICE) CS Colloquium Series
Speaker: Jianzhu Ma, Project Scientist, Dept. of Medicine, UC San Diego
Where: Luddy Hall, Rm. 3166 (Executive Conference Room)
When: Wednesday, February 20, 2019, 10:00 am
Topic: Interpreting genetic variants using network models of a cell
Abstract: Machine learning in biology faces two challenges: First, most of the models are still “black boxes” and difficult to interpret. Although accurate, they provide no meaningful insights about how their decisions are made and are insufficient in disease studies for which clinicians need to understand the mechanisms underlying the predictions. Second, it is often hard to generalize findings from basic biological contexts, where data are abundant, to clinical contexts, where data are few. In the first part of my talk, I will mainly describe how to build a new generation of interpretable deep learning frameworks by coupling the structure of the neural network with the internal workings of cell. In the second part, I will talk about how to adopt the meta-learning framework to learn a general representation of the molecular features which are more adaptable across variable contexts.
Biography: Jianzhu Ma is a Project Scientist in the Department of Medicine at UC San Diego. He received Bachelor’s and Master’s degrees from Peking University in Computer Science Department, and his Ph.D. from Toyota Technological Institute at Chicago under the supervision of Prof. Jinbo Xu. From 2010 and 2015, while working with Xu, his research direction was to predict protein 3D structures by using machine learning approaches. He developed an integrated protein modelling server called RaptorX (http://raptorx.uchicago.edu/). As of 2018, RaptorX has 50,186 registered users and successfully processed 374,684 tasks. Since Nov 2015, Jianzhu Ma joined Prof. Trey Ideker’s lab at UC San Diego as a Project Scientist and focused on developing interpretable and transferable deep learning models for various biological applications, such as therapeutic response of tumor cells, signaling pathway and yeast genetics.
- Wednesday February 20, 2019 10:00 AM
- Wednesday February 20, 2019 11:00 AM
- Luddy Hall
- Yuzhen Ye
- Contact Email