Amino acid biophysical properties in the statistical prediction of peptide-MHC class I binding.

Again the reason might be that as RF and bagging does not use the metric values of the properties, the extra information on partitions provided by including any new property may not be significant, where as for SVM the increment is more significant. • Within the given set of properties SVM achieves the lowest misclassification error with 5 properties (Hydrophobicity, Volume, Isoelec, Branch, Aromatic, where the last two variables are indicator of whether ....), as well as among all 3 property based classifiers SVM achieves the lowest misclassification error of 0.125781, using Hydrophobicity, Volume and Isoelectric point.

also in 2,774 other papers