Hunag, C.-C.C.-C.HunagTu, S.-H.S.-H.TuLien, H.-H.H.-H.LienHuang, C.-S.C.-S.HuangERIC YAO-YU CHUANGLIANG-CHUAN LAI2020-07-012020-07-0120109781424483044https://www.scopus.com/inward/record.uri?eid=2-s2.0-79952028829&doi=10.1109%2fBIBMW.2010.5703853&partnerID=40&md5=5d7181dee2f42af0ce5acc9e710f7278https://scholars.lib.ntu.edu.tw/handle/123456789/507781The study was to compare principle component (PC) versus partial least square (PLS) regression, the former unsupervised and the latter supervised gene component analysis, for highly complicated and correlated microarray gene expression profile. Projection of derived classifiers into independent samples for clinical phenotype prediction was evaluated as well. Previous studies had suggested that PLS might be superior to PC regression in the task of tumor classification since the covariance between predictive and respondent variables was maximized for latent factor extraction. We applied both algorithms for classifier construction and validated their prediction performance on independent microarray experiments. The statistical strategy could reduce high-dimensionality of microarray features and avoid the collinearity problem inherited in gene expression profiles. Proposed predictive model could discriminate breast cancers with positive and negative estrogen receptor status successfully and was feasible for both Taiwanese and Chinese females, both with the same Han Chinese ethnic origin. ?2010 IEEE.[SDGs]SDG3[SDGs]SDG10Breast Cancer; Collinearity; Component analysis; Estrogen receptor; Gene expression profiles; Latent factor; Microarray experiments; Microarray gene expression; Partial least squares; Partial least-square regression; Prediction performance; Predictive models; Principle component; Principle component regression; Statistical strategies; Tumor classification; Bioinformatics; Classifiers; Principal component analysis; Regression analysis; Gene expressionconference paper10.1109/BIBMW.2010.57038532-s2.0-79952028829