SiRPipe: a data mining pipeline to improve ab initio prediction of bacterial small RNAs

This work is focused on ab initio prediction of bacterial small RNAs (sRNAs). We gathered a series of classification features used by existent computational tools in order to select the best ones for classifying putative sRNA sequences, leading to 22 relevant and non-redundant features. The Random Forest algorithm was chosen to build a predictive model using a dataset made with experimentally-validated sRNAs from different substrains. Our approach could achieve a higher predictive power in comparison with current methods for non-coding RNA prediction. We also describe our work in progress comprising a framework for genome-wide sRNA ab initio prediction. Our classifier and datasets can be found at www.sirpipe.com.