Heuristics and meta-heuristics for one-way clustering of gene expression data

One-way gene expression clustering problem consists of identifying or grouping/partitioning characteristic vectors of all related genes. Application of crossing minimization heuristics with recursive noise removal in the domain of unsupervised clustering of one-way gene expression data is a novel idea, presented first by [1]. In this paper, we have further explored this idea through an objective comparison of five crossing minimization heuristics along with two Meta combinations. c have established, after performing extensive computational experiments that for weak clusters even under a noiseless environment the performance of Median Heuristic of [2] drastically deteriorates thus making recursive noise removal ineffective, however MaxSort heuristic of [3] and its derivative gives near perfect results but deteriorates when noise is added. Surprisingly the performance of meta-heuristic i.e. MaxSort followed by Median Heuristic results in less overall time and significantly better overall clustering results. Lastly we apply the heuristics and meta-heuristics on real data and objectively compare the outputs with promising results.