Investigation and comparison of the preprocessing algorithms for microarray analysis for robust gene expression calculation and performance analysis of technical replicates
Proceedings of the IEEE 14th Signal Processing and Communications Applications Conference, SIU 2006
Item Usage Stats
Preprocessing of microarray data involves the necessary steps of background correction, normalization and summarization of the raw intensity data obtained from cDNA or oligo-arrays before statistical analysis. Several algorithms, namely RMA, dChip, and MAS5 exist for the preprocessing of Affymetrix microarray data. Previous studies have identified RMA as one of most accurate algorithms while MAS5 was characterized with lower accuracy and sensitivity levels. In this study, performance of different preprocessing algorithms have been compared in terms of ROC characteristics of pairwise intensity differences of microarray replicates. Our findings indicated that all three algorithms predicted in similar order the quality of the technical replicates obtained from a selected set of latin square experiments . On the other hand, RMA exhibited higher performance in terms of accuracy by maximizing the area under the receiver operating curve. The proposed method also is useful for detection of global and/or local artifacts associated within the technical replicas of a microarray experiment. Therefore this study is unique in the sense that it provides an extensive investigation and comparison of preprocessing algorithms and proposes a novel method for the detection and identification of fine technical replicate pair.
Receiver operating curves