일시: 11/18(목) 16시
연사: 김상수(Sangsoo Kim) 숭실대학교 생명정보학과(Dept. of Bioinformatics and Life Science, Soongsil U.)
제목: Application of sPLS-DA in Clinical Outcome Prediction based on RNA-seq Data
내용:
An RNA-seq study of clinical samples is useful in discovering a panel genes for the prediction of clinical outcomes. As such, it is critical to select the smallest possible number of features that can collectively achieve high accuracy. Partial least squares discriminant analysis (PLS-DA) is particularly useful in that it is robust to the cases where many features are not independent from each other, which is the case for gene expression profiles. PLS-DA can deal with multi-class problems easily. Moreover, it has a sparse version (sPLS-DA), which allows the user tweaking the number of genes to select. In this talk, I will start by presenting an example that has been successfully dealt with sPLS-DA, and finish up with a short course on its algorithm.