Skip to Main Content
In gene expression data, a bicluster is a subset of genes exhibiting a consistent pattern over a subset of the conditions. In this paper, we propose a new method to detect biclusters in gene expression data. Our approach is based on the high dimensional geometric property of biclusters and it avoids dependence on specific patterns, which degrade many available biclustering algorithms. Furthermore, we illustrate that a biclustering algorithm can be decomposed into two independent steps and this not only helps to build up a hierarchical structure but also provides a coarse-to-fine mechanism and overcome the effect of the inherent noise in gene expression data. The simulated experiments demonstrate that our algorithm is very promising.
Machine Learning and Cybernetics, 2005. Proceedings of 2005 International Conference on (Volume:6 )
Date of Conference: 18-21 Aug. 2005