Genome-scale MicroRNA target prediction through clustering with Dirichlet process mixture model

Abstract

BackgroundMicroRNA regulation is fundamentally responsible for fine-tuning the whole gene network in human and has been implicated in most physiological and pathological conditions. Studying regulatory impact of microRNA on various cellular and disease processes has resulted in numerous computational tools that investigate microRNA-mRNA interactions through the prediction of static binding site highly dependent on sequence pairing. However, what hindered the practical use of such target prediction is the interplay between competing and cooperative microRNA binding that complicates the whole regulatory process exceptionally.ResultsWe developed a new method for improved microRNA target prediction based on Dirichlet Process Gaussian Mixture Model (DPGMM) using a large collection of molecular features associated with microRNA, mRNA, and the interaction sites. Multiple validations based on microRNA-mRNA interactions reported in recent large-scale sequencing analyses and a screening test on the entire human transcriptome show that our model outperformed several state-of-the-art tools in terms of promising predictive power on binding sites specific to transcript isoforms with reduced false positive prediction. Last, we illustrated the use of predicted targets in constructing conditional microRNA-mediated gene regulation networks in human cancer.ConclusionThe probability-based binding site prediction provides not only a useful tool for differentiating microRNA targets according to the estimated binding potential but also a capability highly important for exploring dynamic regulation where binding competition is involved.Electronic supplementary materialThe online version of this article (10.1186/s12864-018-5029-7) contains supplementary material, which is available to authorized users.

Covering Hierarchical Dirichlet Mixture Models on binary data to enhance genomic stratifications in onco-hematology

“…Their usage already proved to boost the definition of new clinical classification systems and the discovery of unknown groups of co-occurrent genomic alterations. The HDMMs have also been used transversally on genomic data, including copy number variation plus transcriptomic integration [ 8 ], pan-cancer proteomic characterization [ 9 ], cancer subtyping with microRNA [ 10 ] and disease classes discrimination based on genomics, transcriptomics and epigenomics [ 11 ].…”

Infinite mixtures of multivariate normal-inverse Gaussian distributions for clustering of skewed data

“…Hence, the number of component is inferred simultaneously during parameter estimation. The Dirichlet process mixture models have been applied in a wide variety of applied problems (Lartillot and Philippe, 2004;Huelsenbeck and Andolfatto, 2007;Onogi et al, 2011;Hakguder et al, 2018). Early research on developing the Dirichlet process mixture models goes back to three decades (West, 1992;Escobar and West, 1995;Maceachern and Müller, 1998).…”