Classification and class prediction for different chemical structures using gene expression data

Abstract

For years microarray-based classification has been a major topic in statistics, bioinformatics and biomedical research but because of the large number of variables as compared to the sample size, traditional statistical methods have been unsatisfactory. Thus, special methods for microarray data together with data mining technologies have been developed to address these unsatisfactory issues. The aim of this project was to apply some of these technologies to build classification function(s) that can be used to classify chemical compounds used in the early stage of drug discovery, to their identified clusters and also to predict the cluster of a new chemical compound based on gene expression data. The data used contained sixty chemical compounds grouped into three clusters GC14, GC22 and GC29 with the clusters having 32, 13 and 15 chemical compounds respectively and a total of 7722 genes retained after normalization and gene filtering.