Table 1 Predictive performance across training and non-redundant blind test sets.

From: Prediction of rifampicin resistance beyond the RRDR using structure-based machine learning approaches

Metric

Sensitivity

Specificity

MCC

F1 score

Accuracy

Precision

All

Training (n = 231)

89.7%

100%

0.72

0.90

90.9%

100%

Test (n = 88)

100%

57.1%

0.71

0.91

89.8%

88.2%

Non-RRDR

Training (n = 193)

87.9%

100%

0.72

0.94

89.6%

100%

Test (n = 39)

100%

76.2%

0.77

0.88

87.2%

78.3%

  1. Numbers show performances throughout the whole gene, and outside of the rifampicin resistance determining region (RRDR). Performance on mutations within the RRDR was equivalent, with 100% sensitivity and 0% specificity (no susceptible mutations present).
Back to article page