Table 3 The performance of SVM only trained on the key codons. The SVM models were trained using FCU and RSCU values of the codons listed in Figure 2 as sequence attributes respectively. These codons are thought to play important roles in classification.