Related Experiment Videos
Reduced bio basis function neural network for identification of protein phosphorylation sites: comparison with
Emily A Berry1, Andrew R Dalby, Zheng Rong Yang
1Department of Computer Science, School of Engineering, Computer Science and Mathematics, University of Exeter, UK. e.a.berry@exeter.ac.uk
Computational Biology and Chemistry
|March 17, 2004
Summary
This study compares three algorithms for predicting protein phosphorylation sites. The decision tree algorithm C4.5 achieved the highest accuracy, identifying key amino acids important for phosphorylation.
Area of Science:
- Biochemistry
- Bioinformatics
- Computational Biology
Background:
- Protein phosphorylation is a crucial post-translational modification regulating protein function, cell signaling, and apoptosis.
- Predicting phosphorylation sites is challenging due to the broad substrate specificity of protein kinases.
Purpose of the Study:
- To compare the prediction efficiency and knowledge extraction capabilities of three machine learning algorithms: back-propagation neural networks (BPNNs), decision tree algorithm C4.5, and reduced bio-basis function neural network (rBBFNN).
Main Methods:
- Utilized BPNNs, C4.5, and rBBFNN for predicting protein phosphorylation sites.
- Evaluated algorithms based on prediction efficiency, speed, sensitivity, robustness, and accuracy.
- Examined the knowledge extraction capability of each algorithm.
Main Results:
- All three algorithms demonstrated effectiveness in predicting phosphorylation sites.
- rBBFNN exhibited the fastest performance and highest sensitivity.
- BPNN showed the most robust performance with the highest area under the ROC curve.
- C4.5 achieved the highest prediction accuracy and provided insights into important upstream amino acids for serine/threonine and tyrosine phosphorylation.
Conclusions:
- Machine learning algorithms, including BPNN, C4.5, and rBBFNN, are effective tools for predicting protein phosphorylation sites.
- C4.5 offers a balance of high accuracy and valuable insights into phosphorylation site determinants.
- Further investigation into the specific amino acid contexts can enhance understanding of kinase specificity.