BindN for prediction of DNA and RNA binding residues in proteins | ||
BindN applies support vector machines (SVMs)
to prediction of DNA and RNA-binding residues from sequence features,
including the side chain pKa value, hydrophobicity index and molecular
mass of an amino acid. The SVM classifiers have been constructed
using two curated datasets (PDNA-62 and
PRINR25) from the
Protein Data Bank.
For DNA-binding residues, the prediction accuracy estimated from
cross-validation is about 70% with equal sensitivity and specificity.
For RNA-binding residues, the estimated
accuracy is approximately 68%.
Comparable results have also been obtained using the two test datasets,
TestPDB
and TestSP. To reduce the number of false positive predictions, users may
choose a high specificity value and use the DNA or RNA-binding domain
(if known) instead of the full-length protein
sequence as the input to BindN. Please
send your comments or suggestions to
ljwang@ksu.edu. |