201. SusMiRPred: Ab Initio SVM Classification for Porcine MicroRNA Precursor Prediction
- Author
-
De-Li Zhang, Zhen-Hua Zhao, Wen-Qian Zhang, Yang Zhang, Peng-Fang Zhou, and Fei Zhang
- Subjects
Java package ,Computer science ,business.industry ,Molecular biophysics ,Ab initio ,Computational biology ,Machine learning ,computer.software_genre ,Support vector machine ,microRNA ,Artificial intelligence ,Sequence structure ,business ,computer ,Classifier (UML) ,Support vector machine classification - Abstract
MicroRNA (miRNA), which is short non-coding RNA, plays important roles in almost all biological processes examined. Several classifiers have been applied to predict humans, mice and rats precursor miRNAs (pre-miRNAs), but no classifier is applied to classify porcine pre-miRNAs only based on the porcine pre-miRNAs because of little known miRNA component in the porcine genome. Here, we developed a novel classifier, called SusMiRPred, to predicted porcine pre-miRNAs. Trained on 60 porcine pre-miRNAs and 65 pseudo procine hairpins, SusMiRPred achieve 86.4% (5-fold cross-validation accuracy) and 0.9144 (ROC score). Tested on the remaining 14 porcine pre-miRNAs and 1000 pseudo hairpins, it reports 100% (sensitivity), 87.3% (specificity) and 87.5% (accuracy). SusMiRPred was proved an effective ab initio Support Vector Machine (SVM) classifier for predicting porcine pre-miRNAs and encapsulated with a Java package for other users utilizing it expedient. Furthermore, another Java package, called SusMiRFilter, was developed to filter out the short sequences which have not the pre-miRNAs sequence structure features.
- Published
- 2010
- Full Text
- View/download PDF