SCPRED: Accurate prediction of protein structural class for sequences of twilight-zone similarity with predicting sequences
2008

SCPRED: A Method for Predicting Protein Structural Classes

Sample size: 1673 publication Evidence: high

Author Information

Author(s): Kurgan Lukasz, Cios Krzysztof, Chen Ke

Primary Institution: University of Alberta

Hypothesis

Can we improve the prediction accuracy of protein structural classes for sequences with low identity using a new method?

Conclusion

The SCPRED method can accurately predict structural classes for protein sequences that share low identity with known sequences.

Supporting Evidence

  • SCPRED achieved 80.3% accuracy in predicting structural classes.
  • The method was tested on datasets of 1673 protein chains.
  • SCPRED outperformed over a dozen recent competing methods.
  • Features used in SCPRED were based on secondary structure predictions.
  • The method can be used as a post-processing filter to improve other classification methods.

Takeaway

This study created a new tool called SCPRED that helps scientists figure out the structure of proteins even when they don't look very similar to other proteins.

Methodology

The SCPRED method uses a support vector machine classifier with a custom-designed feature vector based on secondary structure predictions.

Limitations

The predictions for mixed structural classes (α+β and α/β) are of lower quality compared to all-α and all-β classes.

Statistical Information

P-Value

p<0.05

Statistical Significance

p<0.05

Digital Object Identifier (DOI)

10.1186/1471-2105-9-226

Want to read the original?

Access the complete publication on the publisher's website

View Original Publication