Skip to main content

Table 2 The strongest rules for highly and low pathogenic viruses from the HA classifier

From: A complete map of potential pathogenicity markers of avian influenza virus subtype H5 predicted from 11 expressed proteins

  Rule Accuracy (%) Support Class-Specific-Coverage (%)
HP-Rules IF P43(HA1) = D THEN virus = HP 99.8 1225 86
IF P83(HA1) = A THEN virus = HP 100.0 807 57
IF P71(HA1) = I THEN virus = HP 100.0 759 53
LP-Rules IF P43(HA1) = S THEN virus = LP 95.2 589 99
IF P83(HA1) = D THEN virus = LP 94.6 571 95
IF P107(HA1) = S THEN virus = LP 95.8 552 93
IF P138(HA1) = N THEN virus = LP 92.7 536 88
IF P309(HA1) = D THEN virus = LP 94.9 533 89
IF P320(HA1) = V THEN virus = LP 95.7 532 90
IF P195(HA1) = N THEN virus = LP 88.8 400 63
IF P16(SP) = G THEN virus = LP 89.3 392 62
IF P203(HA2) = I THEN virus = LP 82.4 380 55
IF P6(SP) = I THEN virus = LP 97.5 354 61
IF P7(SP) = A THEN virus = LP 98.0 352 61
IF P3(SP) = R THEN virus = LP 94.1 341 57
IF P240(HA1) = S THEN virus = LP 95.2 332 56
IF P275(HA1) = D THEN virus = LP 97.3 300 52
  1. Accuracy is the percentage of the sequences in the support set correctly classified by the rule. Support is the number of sequences that satisfy the “IF” conditions of the rule. Class-Specific-Coverage is the percentage per class (i.e. HP or LP, respectively) of the sequences that support the rule and are correctly classified by the rule. For instance, if a rule is an HP class rule then the Class-Specific-Coverage gives the percentage of the HP sequences classified correctly by this rule