Porpoise: computational analysis and prediction of RNA pseudouridine sites by a stacked machine learning framework.
This study proposes a novel bioinformatics approach, termed Porpoise, for accurate identifying RNA pseudouridine sites. Porpoise is developed by comprehensively evaluated 18 popular feature encoding schemes and four types of features, including binary features, pseudo k-tuple composi-tion (PseKNC), nucleotide chemical property (NCP) and position-specific trinucleotide propensity based on single-strand (PSTNPss) are selected and feed into the stacked framework to construct-ing the final meta-learning model.Cross-validation tests on the benchmark dataset and independent tests demonstrate that Porpoise achieved superior predictive performance than state-of-the-art approaches.
|