A machine learning approach to DNA microarray biclustering analysis

Based on well-established machine learning techniques and neural networks, several biclustering algorithms can be developed for DNA microarray analysis. It has been recognized that genes (even though they may belong to the same gene group) may be co-expressed via a diversity of coherence models. One convincing argument is that a gene may participate in multiple pathways that may or may not be co-active under all conditions. It is biologically more meaningful to cluster both genes and conditions in gene expression data - leading to the so-called biclustering analysis. In addition, we have developed a set of systematic preprocessing methods to effectively comply with various coherence models. This paper will show that the proposed framework enjoys a vital advantage of ease of visualization and analysis. Because a gene may follow more than one coherence models, a multivariate biclustering analysis based on fusion of scores derived from different preprocessing methods appears to be very promising. This is evidenced by our simulation study. In summary, this paper shows that machine learning techniques offers a viable approach to identifying and classifying biologically relevant groups in genes and conditions.