Clustering Analysis Of MiRNA Based On KK-Means Algorithm

Many studies have shown that regulating is the main way of participating in the activities of life and its abnormal behavior has heavily influence on life.There are two kinds of reality to make people have an in-depth study on miRNA.First,the different regulating behavior of miRNA is the key factor causing many diseases.Thus,the study about commonness and individuality of miRNA is urgently needed through existing and little miRNA information.Second,the number of known,associated miRNA-disease correlation is very few.The research about relationship between miRNA making use of the similar relationship network of the disease has becoming a research hot spot in nowadays.The proposed KK-Means algorithm is used to make clustering analysis of miRNA based on a miRNA matrix combining the known,associated miRNA-disease correlation with the similarity network of diseases in this paper.The paper mainly works as follows:A miRNA matrix is constructed by using the associated miRNA-disease network on the basis of K nearest neighbor.However,the miRNA matrix usually is a sparse matrix and do not carry out enough information because the number of the associated miRNA-disease is few in fact.Thus,calculating the related degree of the unknown miRNA-disease and repairing the miRNA matrix are essential combining with the similarity network of diseases involved in miRNA-disease to have the miRNA matrix contain enough information.Some analysis work of classic K-Means algorithm is done and the deficiency of it is listed.The paper explains why it has these weaknesses.Then,on the basis of fully study of the existing clustering algorithms,some methods of overcoming these weaknesses and the reason of why do in that manner are presented.Finally,a modified K-Means algorithm is proposed based on the above analysis.Experiments on the real data sets have shown that the modified algorithm have better performance that classic K-Means algorithm.The paper applies the modified algorithm on the miRNA matrix to cluster.In the last,this paper makes a system design and implement of the modified algorithm and other K-Means algorithm to handily cluster and make a comparison among different clustering algorithms.Using the visualization technology is used to present the clustering result and deepen the understanding of clustering result produced by different clustering algorithms.