Table 3.

Evaluation results of miRNA identification. The evaluation results of miRNA identification using K-means clustering algorithm for the three topological index families are shown. In this application case study, the K-means algorithm is run independently for 50 times. For each test, 200 real pre-miRNAs are randomly chosen from the 1,082 miRNAs in dataset of Table 1 and the corresponding 1,000 pseudo pre-miRNAs are generated as reference set. The clustering accuracy, sensitivity and specificity are employed here to evaluate the performance of the identification results for Wiener indices, Balaban indices and Randić indices, respectively.

Index	Clustering accuracy			Sensitivity			Specificity
	Mean ± SD	Min	Max	Mean ± SD	Min	Max	Mean ± SD	Min	Max
Balaban	0.9534 ± 0.005	0.9475	0.9658	0.9597 ± 0.0072	0.955	0.980	0.9621 ± 0.0068	0.953	0.968
Wiener	0.968 ± 0.0014	0.9658	0.9708	0.9623 ± 0.0026	0.957	0.985	0.9891 ± 0.002	0.986	0.993
Randić	0.9849 ± 0.0026	0.9808	0.9908	0.9842 ± 0.0047	0.975	0.990	0.9851 ± 0.0033	0.979	0.992