日時 2006年 4月 25日(火) 15時〜16時40分 場所 経済学部新棟3階第2教室 [※ 今年度は教室が変更になっております] 講演者 宮田 敏(財団法人癌研究会ゲノムセンター) 演題 Medical informatics for OMICS data 概要: Recently many powerful technologies for detection of genome related data have been developed and utilized to explore important biomarkers and genes related to treatment response. We discuss about high throughput genome-related data, including single nucleotide polymorphism(SNP), DNA microarray data for gene expression, and protein expression profile data. We also argue about bioinformatics methodologies suitable for analysis of genome-related data. The SNP data characterize genetic constitution in patients and are, for example, used to predict adverse effect of medicine. We describe the association study of SNP data and the haplotype based analysis thought to be more efficient than the individual SNP based approach. The gene and/or protein expression profile data are thought to reflect genetic individuality and are used for personalized diagnosis of patients. We introduce some useful methods, including AdaBoost, to analyze these data. AdaBoost is robust outliers, capable of handling missing values and appropriate for analysis of gene and/or protein expression data. To detect expression signals from protein profile data, it is necessary to denoize row data which is inhomogeneously smooth. To fit a trend line to protein profile data, we introduce adaptive free-knot spline and adaptive model selection criterion(Miyata, S. and Shen, X. (2005)).
Tokyo University