Comprehensive machine-learning-based analysis of microRNA–target interactions reveals variable transferability of interaction rules across species

Table 5 XGBoost performance measurements

Dataset	AUC^a	ACC ^b	TPR ^c	TNR^d	MCC ^e	F1 score
ca1	0.983	0.937	0.932	0.943	0.874	0.937
	(0.001)	(0.002)	(0.004)	(0.004)	(0.004)	(0.002)
ce1	0.955	0.889	0.89	0.889	0.779	0.89
	(0.009)	(0.014)	(0.018)	(0.014)	(0.028)	(0.014)
ce2	0.958	0.891	0.884	0.899	0.783	0.89
	(0.012)	(0.016)	(0.02)	(0.019)	(0.032)	(0.017)
h1	0.908	0.824	0.816	0.833	0.649	0.822
	(0.006)	(0.007)	(0.008)	(0.008)	(0.014)	(0.007)
h2	0.972	0.904	0.886	0.924	0.809	0.902
	(0.003)	(0.007)	(0.012)	(0.011)	(0.014)	(0.007)
h3	0.914	0.835	0.823	0.849	0.671	0.832
	(0.004)	(0.007)	(0.011)	(0.009)	(0.014)	(0.008)
m1	0.914	0.847	0.834	0.862	0.695	0.844
	(0.007)	(0.015)	(0.014)	(0.024)	(0.031)	(0.014)
m2	0.963	0.9	0.891	0.909	0.8	0.899
	(0.002)	(0.004)	(0.003)	(0.005)	(0.008)	(0.004)

The cells contain the means and standard deviations (in brackets) acquired from 20 models that were trained and evaluated on different training-testing dataset splits
^aArea under the receiver operating characteristic curve
^bOverall accuracy
^cTrue Positive Rate (Sensitivity)
^dTrue Negative Rate (Specificity)
^eMatthews correlation coefficient

ISSN: 1471-2105