A Local Genetic Algorithm for the Identification of Condition-Specific microRNA-Gene Modules

Cancer causes millions of death every year and abnormal regulation of gene expression is responsible for the development of cancer. Thus, it is crucial to understand the underlying regulatory mechanisms, which includes two main regulators, microRNA and transcription factor. In this study, we proposed a new computational method to identify the regulatory modules that show the competitive and combinatory effect of microRNAs and transcription factors on the same set of regulated genes. Our method was based on a heuristic optimization approach method that combines specifically designed genetic algorithm and local search for the module identification. Two sets of matched gene expression and microRNA expression profiles, microRNA target predictions and transcription factor target predictions were integrated to achieve the goal. We demonstrated that the method could identify highly scored modules with statistical significance and biological relevance. We further presented literature support for several core regulators and predicted regulations. Our method can detect condition-specific regulatory modules and provide proper ways for evaluation. Despite still existing room for improvement, the pipeline developed from this study can be readily used by clinical researchers.