Multiple gene sets for cancer classification using gene range selection based on random forest

Abstract

The advancement of microarray technology allows obtaining genetic information from cancer patients, as computational data and cancer classification through computation software, has become possible. Through gene selection, we can identify certain numbers of informative genes that can be grouped into a smaller sets or subset of genes; which are informative genes taken from the initial data for the purpose of classification. In most available methods, the amount of genes selected in gene subsets are dependent on the gene selection technique used and cannot be fine-tuned to suit the requirement for particular number of genes. Hence, a proposed technique known as gene range selection based on a random forest method allows selective subset for better classification of cancer datasets. Our results indicate that various gene sets assist in increasing the overall classification accuracy of the cancer related datasets, as the amount of genes can be further scrutinized to create the best subset of genes. Moreover, it can assist the gene-filtering technique for further analysis of the microarray data in gene network analysis, gene-gene interaction analysis and many other related fields. © 2013 Springer-Verlag.