Wednesday, April 13th, 2011
2:45pm - 3:45pm, in McCormack 2-116

Kourosh Zarringhalam

Boston College

MicroRNA classification and integration of chemical footprinting data into RNA secondary structure prediction

Abstract: MicroRNAs (miRNAs) are short (~22 nt) endogenous non-coding RNAs that play an important role in post-transcriptional gene regulation. The miRNA precursor (pre-miRNA) has a characteristic hairpin structure with Boltzmann basepairing probabilities in the ensemble of low energy secondary structures significantly different from those of other similar hairpin structures, allowing us to apply machine learning classification methods to predict miRNAs based on structural features. We assess their discriminatory power by training a Support Vector Machine (SVM) classifier on known miRNAs and protein-coding sequences. 5-fold cross-validation tests yield a high accuracy of 0.95. We apply our classification method to NextGen sequencing of short RNAs from human cell lines to identify more than 100 novel human putative miRNAs. Our work suggests that a large number of miRNAs remain to be characterized in eukaryotic genomes. Further we present a method for improving the accuracy of the RNA secondary structure prediction by integrating chemical footprinting data into the Boltzmann Partition function.