Abstract: This paper addressed the problem of detecting scaling patterns in gene expression data. A mean ratio residue as a merit function for Scaling patterns was presented. Based on the mean ratio residue a biclustering method was proposed under genetic rough framework. In this method, initial population is generated by Heuristic Rough Biclustering Algorithm as well as random choice. And then it adjust the seeds with Genetic Rough Algorithm. We tested this method on yeast expression data. The experimental results show that the Genetic Rough based method well improves the performance of heuristic algorithm and biclusters found on the yeast data are biologically significant using online GO Term Finder.

Key words: Biclustcring, Scaling patterns, Genetic rough algorithm, Gene expression data