Recognizing drosha processing sites by a two-step prediction model with structure and sequence information

Abstract

Drosha is a class of RNase III enzyme plays important roles in the microRNA (miRNA) generation by cleaving primary miRNAs to release hairpin-shaped miRNA precursors. Accurately predicting the Drosha cleavage positions (i.e., processing sites) is helpful for the identification of miRNAs and the understanding of miRNA biogenesis mechanisms. In this study, we presented a Drosha processing site predictor, termed DroshaPSP, with a two-step prediction model by integrating structure and sequence features. Testing results on the Drosophila melanogaster miRNA data showed that DroshaPSP obtained a sensitivity of 0.859, a specificity of 0.999, and a Matthew's Correlation Coefficient of 0.864. We also found that the Shannon entropy is a powerful structure feature for DroshaPSP to distinguish true Drosha processing sites from the nearby pseudo processing sites effectively.