Identifying cancer-associated modules from microRNA co-expression networks: a multiobjective evolutionary approach

MicroRNAs (miRNAs) are a class of very small noncoding RNA molecules. Although they are not directly involved in protein translation process, they indirectly regulate production of proteins by targeting different protein-coding genes or messenger RNAs (mRNAs). Several miRNAs are known to have crucial role in progression of different diseases in the human body such as cancer, diabetes, viral infection and cardiovascular diseases. Therefore, it is very important to understand the regulatory relationship among the genes and miRNAs in order to find the potential drug targets for these life-threatening diseases. In this article, a multiobjective miRNA module detection algorithm has been proposed to identify a group of miRNAs associated with several cancer types. This module detection algorithm optimizes two objective functions simultaneously. The first objective function is based on the change in miRNA co-expression pattern across the different phenotypic conditions, and the second objective function is based on the functional similarity within the miRNA pairs. Here, non-dominated sorting genetic algorithm-II (NSGA-II) has been utilized to optimize both the objective functions simultaneously so that differentially co-expressed miRNA modules having greater functional similarity can be detected. The superiority of the proposed technique is demonstrated by comparing its performance in identifying microRNA markers with that of the other existing module detection algorithms. Furthermore, the biological significance of the mRNA targets of the identified miRNA markers has been investigated.