ㅇ 제목 : Machine Learning for Bio-medical Data Analysis Towards Precision Medicine

ㅇ 일시 : 2018.05.15(화) 오전 10:45

ㅇ 장소 : ETRI 12연구동 301호 회의실

ㅇ 강사 : 이슬 교수(서울대)

ㅇ 요약 : 

Machine learning methods are widely applied in analysis of biomedical data. In this talk, I’ll introduce several machine learning methods that I have worked on with the focus on development and application of Interpretable method, Integrative Method, and Utilization of Prior Knowledge. In the interpretable methods, a somatic mutation profiling method based on non-negative matrix factorization is introduced that is made more interpretable by integrating Gene Ontology terms. In the Integrative analysis, two tensor analysis method introduced for integrative analysis of multi-platform genome data. Tensors, multi-mode arrays, are natural representations of multi-mode data such as various bio-data, such as miRNA, methylation, gene expression, and mutation information of cancer patients. Just as non-negative matrix factorization methods have been used to analyze the uni-mode genome data, tensor decomposition methods can be used to analyze the multi-mode data. First method, SNeCT focus on data scalability and missing data problem. Second method, GIFT, focuses on increasing interpretability of the results by utilizing gene function as optimization constraints. In the utilization of prior knowledge, I look at gene prioritization problem and utilization of protein structural knowledge in generating features for classifying a SNP as benign or deleterious and show how structural prior knowledge aids in determining deleterious effects in rare variant rare disease cases.