miRRim: A novel system to find conserved miRNAs with high sensitivity and specificity

Abstract

The identification of novel miRNAs has significant biological and clinical importance. However, none of the known miRNA features alone is sufficient for accurately detecting novel miRNAs. The aim of this paper is to integrate these features in a straightforward manner for detecting miRNAs with better accuracy. Since most miRNA regions are highly conserved among vertebrates for the ability to form stable hairpin structures, we implemented a hidden Markov model that outputs multidimensional feature vectors composed of both evolutionary features and secondary structural ones. The proposed method, called miRRim, outperformed existing ones in terms of detection/prediction performance: The total number of predictions was smaller than with existing methods when the number of miRNAs detected was adjusted to be the same. Moreover, there were several candidates predicted only by our method that are clustered with the known miRNAs, suggesting that our method is able to detect novel miRNAs. Genomic coordinates of predicted miRNA can be obtained from http://mirrim.ncrna.org/.

Genome-wide discovery of miRNAs using ensembles of machine learning algorithms and logistic regression

“…Examples of early machine learning classifiers for miRNA discovery are Triplet-SVM (Xue et al, 2005) and mir-abela (Sewer et al, 2005), which are based on Support Vector Machines (SVMs). Subsequently, miRNA prediction algorithms relying on many different classifiers have been published, including Hidden Markov Models (HMMs) (Terai et al, 2007;Agarwal et al, 2010), random forests (Jiang et al, 2007;Gudyś et al, 2013), artificial neural networks (Rahman et al, 2012) and decision trees (Tyagi et al, 2008). The emergence of next-generation sequencing (NGS) technologies led to development of prediction methods using read mapping in genomes.…”

Direct Updating of an RNA Base-Pairing Probability Matrix with Marginal Probability Constraints

“…The base-pairing probability matrix (BPPM) of an RNA sequence, which stores all the probabilities for possible base pairs in an RNA sequence (see Section 2 for the detailed definition), has played an essential role in a number of algorithms in RNA informatics, including RNA (common) secondary structure predictions (Seemann et al, 2008,) Lu et al, 2009;Hamada et al, 2009bHamada et al, , 2011cSato et al, 2011, multiple alignment of RNA sequences (Hofacker et al, 2004;Katoh and Toh, 2008;Hamada et al, 2009a;Sahraeian and Yoon, 2011), RNA-RNA interaction predictions (Kato et al, 2010;Seemann et al, 2011), RNA motif search (Hamada et al, 2006), and miRNA gene finding (Terai et al, 2007) (Wei et al, 2011). Estimating accurate base-pairing probabilities, therefore, has the potential to improve those algorithms without modifying them.…”

Finding Cancer-Associated miRNAs: Methods and Tools

“…These studies, however, did not integrate conservation information in their algorithms, an important feature of the majority of miRNA genes. More recently two computational tools miRRim [30] and SSCprofiler [31] also employing HMMs proved to be very effective, achieving high performance on identifying miRNAs in the human genome.…”

“…These methods are exceptionally useful as they produce large datasets that offer a relatively accurate expression map for small RNAs in the genome. However, since largescale expression data are usually limited by the specific tissue and developmental stage of their samples, only the coupling of such data to computational tools (as done in two recent studies [30] [31]) can facilitate rapid and precise detection of novel miRNAs, while at the same time giving greater credence to computational predictions.…”