Bioinformatics Algorithms And Application For MicroRNA Deep-sequencing Data Analysis

With the development of next generation sequencing (NGS) technique, manysoftware tools were emerged for the discovery of novel microRNA (miRNAs) and foranalyzing the miRNAs expression profiles. In this study, we evaluated eight software toolsbased on their common features and key algorithms. Three deep-sequencing datasetscollected from different species and used to assess the computational time, sensitivity andaccuracy of detecting known miRNAs as well as their capacity for predicting novelmiRNAs. Our results provide useful information for researchers to facilitate their selectionof the optimal software tools for miRNA analysis depending on their specific requirements,i.e. novel miRNAs discovery or miRNA expression profile analysis of sequencing datasets.Research method is using three deep-sequencing data sets of C. elegans, G. gallusHomo sapiens. The operating system was Ubuntu8.04.4with version of X₈664bits. Theformula of Sensitivity is: Sensitivity (Sen)=TP/(TP+FN). The formula of Accuracy is:Accuracy (Acc)=TP/(TP+FP+FN). The C. elagans, G. gallus and H. sapiens sequencesand their gene annotation were retrieved from UCSC Genome Browser. For predicting thesecondary structure, mfold was utilized.In terms of three deep-sequencing data sets, miRExpress and mireap took lesscomputational time whereas miRDeep and MIReNA took longer computational time. Interms of the ability of predicting known miRNA, two aspects can be analyzed, one issensitivity, and the other is accuracy. In sensitivity, miRExpress and DSAP reached thehighest scores in predicting C. elagans. miRExpress, DSAP and mirTools had the bestperformance in predicting G. gallus, while miRExpress had the best performance in predicting H. sapiens．For three species, the Venn diagrams of software tools miRDeep, mireap andMIReNA on predicting known miRNAs are shown as follows: they had higher overlap inthe intersection of predicting C. elegans, and comparatively lower in G. gallus and H.sapiens. The predicting known miRNAs were more gathered in C. elegans while they weremore discrete in H. sapiens by these three programs.In terms of predicting novel miRNAs, if it is nematode, MIReNA can be the firstchoice, and then is mireap and miRDeep, if it is vertebrate, mireap can be the first choice,and then is miRDeep and miRTRAP, if it is mammal, MIReNA can be the first choice, andthen is miRDeep or mirTools.As for the miRNA deep-sequencing data analysis; therefore it is a pendent questionto select a "best" tool. Before select "suitable" software tools, the organisms for distinctsoftware tool or gene annotation format must be available. miRExpress can be used in anyorganisms, but mfold and RNALogo have to be used in discovery of novel miRNAs.DSAP can be used for have comparative miRNAomics. Generally appropriate softwaretool should be selected rely on the input and output requirement.