Spring 2020 Seminar: Application of deep neural network methods in genomics research
Seminar | February 19 | 12-1 p.m. | 106 Stanley Hall
Lana Garmire, University of Michigan
Abstract: Population genomics data generally have larger feature sizes than its sample sizes, posing challenges for deep-learning application in this field. We have developed a multi-omics data integration tool called DeepProg, which uses multiple types of genomics data to predict liver cancer patient survival through a deeplearning and machine learning combined approach. We further demonstrate the utility of these methods on tens of thousands of cancer samples in the cancer genome atlas, with accuracy higher than the state of the art method Similarity Network Fusion (SNF). Moreover, unlike SNF, DeepProg is capable of predicting new patient survival. Lastly, I will demonstrate the advantage of using deep-neural network models to impute gene expression, at single cell RNA-Seq level, with a tool called DeepImpute. These methods, from population to single cell levels, collectively shows the promise of applying deep-learning in the genomics field.