Figure 3

miRPara prediction pipeline. (i) if query sequence is longer than 500 nt it is split into overlapping fragments to prevent formation of long range structures. (ii) secondary structure of each fragment is predicted and if it forms a hairpin structure it is classified as a pre-miRNA candidate and passed to next step. (iii) all possible miRNA that can occur within the putative pre-miRNA are generated and, along with the parent pre-miRNA, are passed in turn to the SVM for classification. (iv) all parameters of the pre-miRNA/miRNA pair are calculated and passed through a filtering step. If the SVM classification is positive and the parameters pass the filtering step, the pair is classified as a true pre-miRNA/mi-RNA pair.