MiRduplexSVM: A High-Performing MiRNA-Duplex Prediction and Evaluation Methodology

Abstract

We address the problem of predicting the position of a miRNA duplex on a microRNA hairpin via the development and application of a novel SVM-based methodology. Our method combines a unique problem representation and an unbiased optimization protocol to learn from mirBase19.0 an accurate predictive model, termed MiRduplexSVM. This is the first model that provides precise information about all four ends of the miRNA duplex. We show that (a) our method outperforms four state-of-the-art tools, namely MaturePred, MiRPara, MatureBayes, MiRdup as well as a Simple Geometric Locator when applied on the same training datasets employed for each tool and evaluated on a common blind test set. (b) In all comparisons, MiRduplexSVM shows superior performance, achieving up to a 60% increase in prediction accuracy for mammalian hairpins and can generalize very well on plant hairpins, without any special optimization. (c) The tool has a number of important applications such as the ability to accurately predict the miRNA or the miRNA*, given the opposite strand of a duplex. Its performance on this task is superior to the 2nts overhang rule commonly used in computational studies and similar to that of a comparative genomic approach, without the need for prior knowledge or the complexity of performing multiple alignments. Finally, it is able to evaluate novel, potential miRNAs found either computationally or experimentally. In relation with recent confidence evaluation methods used in miRBase, MiRduplexSVM was successful in identifying high confidence potential miRNAs.

Genome-wide Prediction of microRNAs in Zika virus Genomes Reveals Possible Interactions with Human Genes Involved in the Nervous System Development

“…With the purpose of extracting mature miRNA: miRNA* duplexes from pre-miRNA hairpins, we employed MiRduplexSVM [32]. This approach uses a novel SVM-based methodology and takes into account several aspects of the biogenesis of miRNAs, whereby a duplex is formed before the mature molecule is selected [32].…”

Emerging applications of read profiles towards the functional annotation of the genome

“…Most of the initial efforts for computational prediction of miRNA utilized characteristic hairpin secondary structure of miRNA with homology search (Wang et al, 2005 ; Dezulian et al, 2006 ) or evolutionary conservation (Lai et al, 2003 ; Lim et al, 2003 ). Also methods based on phylogenetic shadowing (Berezikov et al, 2011 ), neighbor step loop search (Ohler et al, 2004 ), minimal folding free energy index (Zhang et al, 2006 ), machine learning (Oulas et al, 2009 ; Karathanasis et al, 2015 ), and statistical approaches (Gkirtzou et al, 2010 ; Karathanasis et al, 2014 ) have been developed. A major drawback of these methods is that they require that the novel miRNAs should either share similar sequence (homology based method) or certain characteristic features (for statistical and machine learning methods) with already known miRNAs.…”

miRLocator: Machine Learning-Based Prediction of Mature MicroRNAs within Plant Pre-miRNA Sequences

“…However, this type of experimental method remains time-consuming because it requires the identification of expressed miRNAs from millions of sequencing reads and has a limited ability to detect miRNAs that exhibit low, linkage,stress, developmental and/or cell-specific expression[ 5 ]. Therefore, experimental technologies must be complemented with computational approaches to identify miRNAs at the genome scale, regardless of the availability of NGS sequencing data [ 5 , 6 ].…”

“…The prediction of mature miRNAs involves determination of the location of mature miRNAs within pre-miRNA sequences. The existing computational approaches for miRNA prediction can be broadly separated into two categories: rule-based approaches[ 8 , 9 ] and machine learning (ML)-based approaches[ 5 , 6 , 9 – 13 ].…”

“…ML was recently applied to the identification of mature miRNAs through the integration of thousands of sequence and structural properties extracted from miRNA sequences and/or miRNA duplexes. Notable ML-based miRNA predictors include miRdup[ 5 ], miRduplexSVM[ 6 ], miRAlign[ 9 ], MatureBayes[ 10 ], MaturePred[ 13 ] and ProMiR[ 14 ]. Although the application of ML technologies has resulted in a marked improvement in the miRNA prediction accuracy relative to that obtained with rule-based approaches, several problems in ML-based miRNA prediction remain unsolved.…”