Prediction Of Cytokine And MiRNA

As the wide application of next-generation sequencing techniques,protein sequences have been rapidly accumulated in last few decades.Specific proteins play critical roles in biological functions and are closely related with human diseases.The upcoming problem is how to accurately identify specific proteins from such large-scale un-characterized proteins.In recent years,computational methods have been developed for the specific protein identification.In particular,many research focuses are on the development of machine learning based methods,which have proven to provide the fast,accurate,and robust identification of specific proteins.In this study,we attempt to enhance the performance of identification methods from the two following factors:(1)feature representation and(2)classifier selection.For feature extraction,we first fuse multiple types of features showing good performance to classify cytokines from non-cytokines,and then employed two feature selection techniques,Max-Relevance-Max-Distance(MRMD),to yield the optimal feature representations.The performance of our method was compared with other methods.For classifier selection,various powerful classifiers are performed,and the one with the highest performance is determined to build the classification model for our method.In this study,we employedback propagationneural network together with 98-dimensional novel features for microRNA precursor identification.Results show that the precision and recall of our method are 96.00% and 96.67%,respectively.Results further demonstrate that the total prediction accuracy of our method is nearly 13.64% greater than the state-of-the-art microRNA precursor prediction software tools.We developed a feature set for cytokine identification,a total of 144 experiments were conducted.Based on the analysis,we learned that our feature sets stably maintain high performance with any of the classifier we used.Finally,we combined 2 classifiers with 3 types of feature sets.The overall performances of the combinations were in the following order from best to worst: 473D+libSVM_Fscore,MRMD+libD3C,and PCA+libSVM_Fscore.