Simple Methods for Peak and Valley Detection in Time Series Microarray Data

Given a set of gene expression time series obtained by a microarray experiment, this work proposes a novel quality control procedure that exploits six analytical methods, each of which allows for the identification in an automated way of genes that have expression spikes within narrow time-windows and over a chosen amplitude threshold. The output of these methods, suitably combined in an automated way, provides an exhaustive list of genes and time points in which abrupt variations have been detected. The quality control on these genes is then performed by a biologist, who classifies the spikes either as biologically relevant or as artifacts. In the latter case, spikes must be eliminated by a smoothing procedure. In this chapter, we first describe the six methods and their iterative and automated implementation. As a case study, we discuss the application of the panel of these six methods to the transcriptome of Plasmodium falciparum intraerythrocytic developmental cycle. Assuming that spikes detected in this set have been labeled as artifacts by a biologist, in the second part of the chapter we discuss the effect of our smoothing procedure for different types of data analysis.