Association between two-component systems gene mutation and Mycobacterium tuberculosis transmission revealed by whole genome sequencing

Table 4 The performance of various models for discriminating cross-regional from within-regional in the lineage4 cohort

Parameters	Training set (n = 3201, 245 cross-regional strains,2956 within-regional trains)		Test set (n = 1373, 93 cross-regional strains, 1280 within-regional strains)
	Random Forest	Gradient Boosted Classification Tree	Random Forest	Gradient Boosted Classification Tree
Kappa	0.649	0.553	0.472	0.435
AUC (95% CI)	0.954 (0.947, 0.961)	0.941 (0.933, 0.949)	0.927 (0.913, 0.941)	0.922 (0.908, 0.936)
Sensitivity (95% CI)	0.981 (0.976, 0.986)	0.458 (0.441, 0.475)	0.971 (0.962, 0.980)	0.363 (0.338, 0.388)
Specificity (95% CI)	0.981 (0.976, 0.986)	0.990 (0.987, 0.993)	0.971 (0.962, 0.980)	0.984 (0.977, 0.991)
PPV (95% CI)	0.732 (0.717, 0.747)	0.783 (0.769, 0.797)	0.543 (0.517, 0.569)	0.649 (0.624, 0.674)
NPV (95% CI)	0.969 (0.963, 0.975)	0.958 (0.951, 0.965)	0.962 (0.952, 0.972)	0.951 (0.940, 0.962)
PLR (95% CI)	23.808 (23.802, 23.814)	18.728 (18.721, 18.735)	14.323 (14.315, 14.331)	13.142 (13.131, 13.153)
NIR (95% CI)	0.042 (-0.017, 0.101)	0.053 (-0.014, 0.120)	0.070 (0.006, 0.134)	0.076 (0, 0.152)
Accuracy (95% CI)	0.954 (0.947, 0.961)	0.951 (0.944, 0.958)	0.937 (0.924, 0.950)	0.938 (0.925, 0.951)

AUC, area under the curve; PPV, positive predictive value; NPV, negative predictive value; PLR, positive likelihood ratio; NLR, negative likelihood ratio; CI, confidence

ISSN: 1471-2164