Predicting DNA-Binding Proteins with SVM
Author Information
Author(s): Kumar Manish, Gromiha Michael M, Raghava Gajendra PS
Primary Institution: Institute of Microbial Technology
Hypothesis
Can support vector machines (SVM) effectively predict DNA-binding proteins using evolutionary profiles?
Conclusion
A highly accurate method has been developed for predicting DNA-binding proteins using SVM and PSSM profiles.
Supporting Evidence
- The SVM models achieved an accuracy of 86.62% using PSSM profiles.
- Models trained on DNAset performed better than those trained on DNAaset.
- The study demonstrated that evolutionary information significantly improves prediction accuracy.
Takeaway
Scientists created a computer program that helps find proteins that can bind to DNA, which is important for understanding how genes work.
Methodology
SVM models were developed using amino acid and dipeptide compositions, as well as PSSM profiles from PSI-BLAST.
Potential Biases
Potential overfitting due to the use of non-redundant datasets.
Limitations
The models trained on DNA domains are not suitable for predicting full-length DNA-binding proteins and vice versa.
Statistical Information
P-Value
p<0.05
Statistical Significance
p<0.05
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website