Related Experiment Videos
Predicting protein-binding regions in RNA using nucleotide profiles and compositions.
Daesik Choi1, Byungkyu Park1, Hanju Chae1
1Department of Computer Science and Engineering, Inha University, Incheon, 22212, South Korea.
BMC Systems Biology
|April 1, 2017
Summary
This study introduces a new support vector machine (SVM) model to accurately predict protein-binding regions in RNA sequences, outperforming existing methods. The developed computational tool offers a powerful approach for analyzing protein-RNA interactions.
Area of Science:
- Computational biology
- Bioinformatics
- Molecular biology
Background:
- Protein-RNA interactions are crucial in biological processes, with increasing data necessitating advanced computational prediction methods.
- Existing computational tools primarily focus on predicting RNA-binding sites in proteins, not protein-binding sites in RNA, which is a more complex challenge.
- Current methods for predicting protein-binding sites in RNA have practical limitations.
Purpose of the Study:
- To develop and evaluate a novel computational model for predicting protein-binding regions within mRNA sequences.
- To address the limitations of existing methods by focusing on predicting binding sites in RNA molecules.
- To provide a more accurate and practical tool for analyzing protein-RNA interactions.
Main Methods:
- Development of a support vector machine (SVM) model utilizing sequence profiles based on log-odds scores of mono- and di-nucleotides, along with nucleotide compositions.
- Rigorous evaluation using 10-fold cross-validation, leave-one-protein-out (LOPO) cross-validation, and independent testing.
- Testing on datasets with varying ratios of binding to non-binding regions to assess model robustness.
Main Results:
- The SVM model achieved high performance on balanced datasets, with 10-fold cross-validation yielding 92.0% accuracy and a Matthews correlation coefficient (MCC) of 0.840.
- Leave-one-protein-out cross-validation demonstrated robust performance with 87.6% accuracy and 0.752 MCC.
- Independent testing achieved 82.2% accuracy and 0.656 MCC, and the model outperformed other state-of-the-art methods on a common dataset.
Conclusions:
- Sequence profiles derived from log-odds scores of nucleotides are highly effective features for predicting protein-binding regions in RNA.
- While nucleotide compositions offer a slight performance enhancement when combined with sequence profiles, the latter are the primary drivers of predictive power.
- The developed model shows significant potential as a powerful predictor of protein-binding regions in RNA, with the program and data publicly available.