Local sequence and structure features in long RNAs

This work proposes an automatic method to analyze different data sets based on their features. The features of human lincRNA were compared to those of mRNA. The features include sequences compositions, structures, and base-pairs probabilities. Afterward, a generic Background Model (BGM) was computed, which fitted the values of these features to a selection of distributions. The fitted distribution is used as a basis to determine cutoff values for significant regions within the sequences for different significant levels. With the help of the BGM, significant regions of lincRNA sequences could be highlighted and compared between different features. Moreover, it was used to compare the similarity of the feature distributions in lincRNA with those of the different parts of the mRNA, namely the 5’UTR, the coding sequence, and the 3’UTR. Finally, each step in our approach was visualized in automatic graphs to enable an easier identification of regions of interest or to manually check the computed results.