miPIE: NGS-based Prediction of miRNA Using Integrated Evidence

Abstract

Methods for the de novo identification of microRNA (miRNA) have been developed using a range of sequence-based features. With the increasing availability of next generation sequencing (NGS) transcriptome data, there is a need for miRNA identification that integrates both NGS transcript expression-based patterns as well as advanced genomic sequence-based methods. While miRDeep2 does examine the predicted secondary structure of putative miRNA sequences, it does not leverage many of the sequence-based features used in state-of-the-art de novo methods.Meanwhile, other NGS-based methods, such as miRanalyzer, place an emphasis on sequence-based features without leveraging advanced expression-based features reflecting miRNA biosynthesis. This represents an opportunity to combine the strengths of NGS-based analysis with recent advances in de novo sequence-based miRNA prediction. We here develop a method, microRNA Prediction using Integrated Evidence (miPIE), which integrates both expression-based and sequence-based features to achieve significantly improved miRNA prediction performance. Feature selection identifies the 20 most discriminative features, 3 of which reflect strictly expression-based information. Evaluation using precision-recall curves, for six NGS data sets representing six diverse species, demonstrates substantial improvements in prediction performance compared to miRDeep2 and miRanalyzer. The individual contributions of expression-based and sequence-based features are also examined and we demonstrate that their combination is more effective than either alone.

Multi-view Co-training for microRNA Prediction

“…This results in a vector of 32 sequenced-based features, pertaining to minimum free energy derived features, sequence/structure triplet features and dinucleotide sequence motifs, and structural robustness features. The eight expression-based features derived in 26 were used as our expression-based feature set, which consist of: (1) percentage of mature paired miRNA nts, (2) number of pairs in lower stem, (3) the percentage of RNA-seq reads in region which are inconsistent with Dicer processing and (4) from the loop region that match Dicer processing, (5) the percentage of RNA-seq reads -(6) RNA-seq-reads from the mature miRNA and -(7) RNA-seq-reads from miRNA* region which match Dicer processing, and (8) the total number of reads in the precursor region, normalized to experiment size. Classification pipeline.…”

“…This creates two views for classification, an expression-based view and a sequence-based view. It should be noted that both of these views have been previously applied to miRNA classification independently 14 and as an integrated feature set 26,27 ; however, multi-view co-training has yet to be explored in the field of miRNA prediction. By applying multi-view co-training, we leverage each view using the other to create iteratively more powerful classifiers.…”

Active and Multi-View Machine Learning for microRNA Prediction

“…Furthermore, this thesis also explores another novel approach to miRNA prediction, which is the combination of expression and sequence-based features for prediction. This idea was first proposed in Robert Peace's PhD thesis [12] and is further explored here and in a separate publication currently under peer review [68]. In this chapter we propose an active learning approach suitable for developing a novel highperformance miRNA prediction tool.…”

“…According to previous work, it is expected that this integrated classifier will show an enhanced performance [12,68] Our proposed active learning algorithm is applied to 6 different species, as specified in Table 2 in Section 3.2. Six active learning models are built for these six species in a similar way to ensure the consistency of our results.…”

“…Amongst all the named tools above, none have considered using both sequence and expression-based features for miRNA prediction. The only method to do so is miPIE [12,68], which takes advantage of the full range of evidence available for classification and prediction by integrating both sets of features. This tool uses an integrated feature set containing 223 sequence and expression-based features, and applies this feature set to six distinct species.…”