Clustering Sleep Deprivation Effects On The Brain Of Drosophila Melanogaster

In almost all successful time-series clustering, feature selection methods are essential. Though lots of feature selection algorithms have being developed from time to time, issues concerning partially unclean data lead to controversies. In this paper, we introduce the idea of selecting interquartile range value (IQR) to set the minimum absolute expression change filtering boundary. The Short Time-Series Expression Miner (STEM) tool is employed to ensure that reliable data which pass the boundary support the STEM clustering method analyses. The main aim of this study is to access as to whether the idea of IQR value setting could ideally compress data and at the same time retrieve the most optimum information. Besides, analysis of clustering profiles generated enable good visualization effect. This is particularly needed in conditional studies of the effect and causality relations as determinant factors to judge the level of correlations among clusters. Our analysis is implemented on a fruit-fly species (Drosophila melanogaster) expression data available from GEO Datasets. The study data consist of 14010 gene profiles recorded in four time points; which subdivided further into three major conditions: unperturbed, perturbed and active.