Mining Association Rules Of Micrornas Based On Information Entropy

With the rapid development of gene sequencing technology, biologiacal data has experienced an exponential growth. The traditional biologiacal methods are limited by some factors, such as the high cost of biological experiments and long cycle. Thus, it is important to accurately identify valuable regulatorary features by a large scale biologiacal data. In recent years, there has been a great success in the research of microRNAs(miRNAs), and miRNAs play an important role in cell’s proliferation, variation, growth, death and disease. Therefore, further studying miRNAs’sequence and structure is helpful for many regulatorary features in the organisms.This paper applies miRNAs association rules based on mutual information and joint information, and there are two different directions between positive and negative regulation of analyzing the role of miRNAs in the organisms. The main studies are listed below:(1)To overcome shortage in the previous studies regarding positive relationship between miRNAs, which ignore miRNAs expression analysis and information exchange between lots of miRNAs in various tissues of the organism, a method identify positive miRNAs association rules using information information, is proposed. Firstly, traditional Apriori algorithm is improved. Support constraint is used to prune the candidate itemsets and finds interesting frequent itemsets. Then, the improved method for calculating confidence in frequent itemsets is employed to extract highly relevant positive association rules. The mature sequence, seed region and secondary structure of miRNAs are used to calculate the average similarity. To find positively related miRNAs which act on the same tissue in the organism, joint entropy and mutual information of information entropy theory is employed to analyze the relationship between miRNAs. Thus, it accurately discovers the positive association rules between miRNAs. Hausdorff distance function is ueded to calculate the similarity of the secondary structure between miRNAs, and geometic average function is employed to calculate the similarity between the mature sequence and seed region. Further, RMS function is used to calculate the overall similarity of structure and sequence.(2)Traditional studies of miRNAs ignore negative relationship in miRNAs expression of various tissues with organisms, and the analysis of information exchange between the miRNAs that inhibite with each other. Therefore, a method, which finds negative miRNAs association rules using information theory, is proposed. Firstly, traditional negative association rules algorithm is adapted. Support constraint is used to prune the infrequent itemsets and find interesting negative itemsets..Then, the improved method for calculating confidence in frequent itemsets is applied to extract highly relevant positive association rules. Pearson correlation coefficient is employed to identify the strong negative correlation itemsets. The improved method for calculating confidence in frequent itemsets is useed to extract highly relevant negative association rules. The overall similarity of miRNAs is calculated by its structure and sequence. To find negatively related miRNAs which act on the same tissue in the organism, joint entropy and mutual information of used to analyze the relationship between miRNAs.The experimental results demonstrate that the proposed methods can find positive and negative associations between miRNAs. It provides more valuable discipline in the diagnosis and treatment of lots of diseases.