Algorithm of RNAcon
RNAcon contains two different prediction models, (i) discrimination between non-coding and coding RNAs and (ii) Classification of predicted ncRNAs into respective classes. In the discrimination between non-coding and coding RNA, we implemented Support Vector Machines (SVMs) based machine learning tool using different nucleotide composition as input features. We used three different datasets: (1) Main dataset, (2) Independent dataset and (3) CONC dataset. Different kernels and parameters of SVM were optimized and models were evaluated using 10-fold cross validation technique. Finally tri-nucleotide compositions (TNC) based model has been implemented in the RNAcon web-server for the discrimination between non-coding and coding RNA sequences.
In the classification mode, we classified the transcripts into 18 different non-coding RNA classes (CD-BOX, HACA-BOX, IRES, LEADER, MIRNA, 5S-rRNA, 5.8S-rRNA, tRNA, 6S-RNA, SRP, tmRNA, Intron-gp-1, Intron-gp-2, SECIS, SSU-rRNA5, T-box, RIBOSWITCH and RIBOZYME). We used previously developed dataset of GraPPLE method (Childs et al 2009). First we predicted the secondary structures of non-coding RNA using IPknot software (Sato et al 2011). The predicted ncRNA structures were used for the calculation of graph properties using igraph R package (Csárdi and Nepusz 2006). Total 20 different graph properties: Articulation points, Average path length, Average node betweenness, Variance of node betweenness, Average edge betweenness, Variance of edge betweenness, Average co-citation coupling, Average bibliographic coupling, Average closeness centrality, Variance of closeness centrality, Average Burt's constraint, Variance of Burt's constraint, Average degree, Diameter, Girth, Average coreness, Variance of coreness, Maximum coreness, Graph density and Transitivity were calculated. The numerical values of these graph properties were used for the different machine learning algorithms of WEKA. WEKA is a single platform of various machine-learning algorithms (Hall et al 2009). We used WEKA 3.6.4 version, which containing different classifiers such as BayeNet, NaiveBayes, MultilayerPerceptron, IBk, libSVM, SMO and RandomForest. We compared classification performance of a variety of algorithms (BayeNet, NaiveBayes, MultilayerPerceptron, IBk, libSVM, SMO and RandomForest). First we trained models on the training dataset using 10-fold cross validation technique and then tested model on the independent (test) dataset.
Algorithm of RNAcon with an example sequence
