Infer genetic/transcriptional regulatory networks by recognition of microarray gene expression patterns using adaptive neuro-fuzzy inference systems

Abstract

In our previous studies, we first noted that dot products of 1st - and 2nd -order differential derivatives are effective to extract paired expression patterns in time-course microarray gene expression data. Here, the feature extraction model and an adaptive neuro-fuzzy inference system are encompassed together for fusing information attained from microarray experiments and known gene-gene interactions confirmed by biological experiments. Having the capability allows computer to associate gene expression patterns with genetic or transcriptional interactions, we can use it to identify other interactions that have not yet been confirmed by biologists. We also bring forward the concept of multilayer adaptive neuro-fuzzy inference systems showing how fuzzy models collaborate with each other on the basis of limited time-course microarray data and take advantage of the known gene-gene interactions. In this chapter, we examine the fuzzy model that may lead to higher true-positive results in prediction of gene networks. We also discuss the relative merits and drawbacks of two methods, and provide recommendations for readers to choose the most suitable method specifically for their applications in bioinformatics. © 2009 Springer-Verlag Berlin Heidelberg.