Predicting Protein Residue Depth Using Sequence Information
Author Information
Author(s): Zhang Hua, Zhang Tuo, Chen Ke, Shen Shiyi, Ruan Jishou, Kurgan Lukasz
Primary Institution: Nankai University, Tianjin, PR China; University of Alberta, Edmonton, AB, Canada
Hypothesis
Can residue depth be accurately predicted from protein sequence using evolutionary information and predicted secondary structure?
Conclusion
The proposed method, RDPred, provides statistically significantly better predictions of residue depth compared to previous methods.
Supporting Evidence
- RDPred outperformed the previous method by Yuan and Wang in predicting residue depth.
- The method showed improved accuracy for both buried and exposed residues.
- Feature selection reduced the dimensionality of the input vector while maintaining prediction quality.
- Hydrophilic and flexible residues were predicted more accurately than hydrophobic and rigid residues.
Takeaway
This study created a new method to guess how deep parts of a protein are buried inside it, which helps understand how proteins work.
Methodology
The method uses support vector regression with features derived from protein sequences, PSI-BLAST scoring matrices, and predicted secondary structures.
Potential Biases
Potential bias in predictions due to the predominance of exposed residues in datasets.
Limitations
The method may not perform as well on datasets with high local sequence similarity.
Statistical Information
P-Value
p<0.00000001
Confidence Interval
(0.037, 1.510)
Statistical Significance
p<0.05
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website