Table 2 The performance of SVM with exclusion of certain amino acids. The SVM models were trained using codon composition, codon bias and amino acid composition as sequence attributes respectively, excluding His, Arg, Asn and Ser from the training datasets.