Accurate classification of immunomodulatory RNA sequences

During the past decade, non-coding RNA has been an exploding area of interest for researches from many fields, from cellular biology to medical applications. A commercial application of these molecules is the RNA interference process (RNAi) from which there are available reagents and a database of documented unwanted side effects, such as the generation of an immunomodulatory response that might be used to enhance the therapeutic effects of gene therapy. The molecules with these properties are denominated immunomodulatory ribonucleotides (IMORNs). We explore here whether or not a recent classification result of these IMORNs [6] can be improved. We focus on a set of sequence-based features to represent the IMORNs and we test different classification algorithms on a dataset of 1122 ssRNA sequences with lengths in the range from 17 to 27 nucleotides proposed in [6]. The best found classifier outperforms the existing results in terms of accuracy, specificity, and Matthews correlation coefficient.