INB-Lunch-Seminar

Biomarker discovery in pooled samples

Anna Telaar, PhD student FBN Dummersdorf

RNA-sample pooling is a common procedure in microarray technology. According to the literature you should avoid pooling for a biomarker study. However sometimes it is necessary, particularly if not enough RNA exists for the hybridization step or because of limited financial resources. Therefore we study pooling effects on several classification methods in microarray experiments. Our simulation framework investigates a two-class classification study to point out how much the outcomes of a single sample and pooling designs differ. The results show how sensitive the effects are in dependency on pool size, discriminating pattern types, number of informative features and the statistical learning method used. The main focus lies on the comparison of important features for the classification for the different designs. Moreover I present our approach to a design issue with experimental data.