iRPIS-PseNNC: identifying RNA-protein interaction sites by incorporating the position-specific dinucleotide propensity into ensemble random forest approach

3rd International Conference on Advances in Energy, Environment and Chemical Engineering

能源学;生态环境科学;化学工业

Li, Long^1 ; Chen, Guojin^1 ; Jin, Tingdu^1

School of Mechanical Engineering, Hangzhou Dianzi University, Hang Zhou

310018, China^1

关键词: Cross validation;    Ensemble classifiers;    Intelligent recognition;    Random sampling;    Reliable recognition;    RNA-protein complexes;    RNA-protein interactions;    Training dataset;   

Others  :  https://iopscience.iop.org/article/10.1088/1755-1315/69/1/012056/pdf
DOI  :  10.1088/1755-1315/69/1/012056

学科分类：环境科学（综合）

来源: IOP

【 摘 要 】

As the pile of RNA-Protein complexes sequences mounted, in order to overcome time-consuming problem of the traditional identify RNA-Protein interaction sites (RPIS) method, it is urgent need develop intelligent recognition approach for quickly and reliable recognition of the RNA-Protein interaction sites (RPIS). To settle the question, we developed a new method named iRPIS-PseNNC, in which each sample is a nineteen nucleotides segment that for positive the centre of the segments is RPIS and for negative the segments centre is non-RPIS, and the sample was obtained by sliding window. The RNA sample was formulated by combining the dipeptide position-specific propensity into random forest approach, and by using the random sampling to balance the training dataset. According the voting system, we combine eleven random forest together to construct an ensemble classifier. It is shown that via the rigorous cross validations that the new predictor "iRPIS-PseNNC" achieved very high percentage of accuracy than any other existing algorithms in this field, indicating that the iRPIS-PseNNC predictor will be an effective tool for prediction RNA-Protein interaction sites.