computational prediction of polycomb-associated long non-coding rnas计算预测polycomb-associated长非编码rna.pdfVIP

Computational Prediction of Polycomb-Associated Long Non-Coding RNAs 1 2 3 Galina V. Glazko *, Boris L. Zybailov *, Igor B. Rogozin 1 Division of Biomedical Informatics, University of Arkansas for Medical Sciences, Little Rock, Arkansas, United States of America, 2 Department of Biochemistry and Molecular Biology, University of Arkansas for Medical Sciences, Little Rock, Arkansas, United States of America, 3 National Center for Biotechnology Information, National Library of Medicine, National Institutes of Health, Bethesda, Maryland, United States of America Abstract Among thousands of long non-coding RNAs (lncRNAs) only a small subset is functionally characterized and the functional annotation of lncRNAs on the genomic scale remains inadequate. In this study we computationally characterized two functionally different parts of human lncRNAs transcriptome based on their ability to bind the polycomb repressive complex, PRC2. This classification is enabled by the fact that while all lncRNAs constitute a diverse set of sequences, the classes of PRC2-binding and PRC2 non-binding lncRNAs possess characteristic combinations of sequence-structure patterns and, therefore, can be separated within the feature space. Based on the specific combination of features, we built several machine-learning classifiers and identified the SVM-based classifier as the best performing. We further showed that the SVM-based classifier is able to generalize on the independent data sets. We observed that this classifier, trained on the human lncRNAs, can predict up to 59.4% of PRC2-binding lncRNAs in mice. This suggests that, despite the low degree of sequence conservation, many lncRNAs play functionally conserved biological roles. Citation: Glazko GV,