Microarray data mining and gene regulatory network analysis
The novel molecular biological technology, microarray, makes it feasible to obtain quantitative measurements of expression of thousands of genes present in a biological sample simultaneously. Genome-wide expression data generated from this technology are promising to uncover the implicit, previously unknown biological knowledge. In this study, several problems about microarray data mining techniques were investigated, including feature (gene) selection, classifier genes identification, generation of reference genetic interaction network for non-model organisms and gene regulatory network reconstruction using time-series gene expression data. The limitations of most of the existing computational models employed to infer gene regulatory network lie in that they either suffer from low accuracy or computational complexity. To overcome such limitations, the following strategies were proposed to integrate bioinformatics data mining techniques with existing GRN inference algorithms, which enables the discovery of novel biological knowledge. An integrated statistical and machine learning (ISML) pipeline was developed for feature selection and classifier genes identification to solve the challenges of the curse of dimensionality problem as well as the huge search space. Using the selected classifier genes as seeds, a scale-up technique is applied to search through major databases of genetic interaction networks, metabolic pathways, etc. By curating relevant genes and blasting genomic sequences of non-model organisms against well-studied genetic model organisms, a reference gene regulatory network for less-studied organisms was built and used both as prior knowledge and model validation for GRN reconstructions. Networks of gene interactions were inferred using a Dynamic Bayesian Network (DBN) approach and were analyzed for elucidating the dynamics caused by perturbations. Our proposed pipelines were applied to investigate molecular mechanisms for chemical-induced reversible neurotoxicity.