A genotypic method for determining HIV-2 coreceptor usage enables epidemiological studies and clinical decision support

Table 1 Classifier AUCs per run of cross validation

CV Run	RBF (σ = 0.001)	Linear	Polynomial (degree = 2)	Edit Kernel (γ = 0.005, PAM70)
1	0.9475	0.9459	0.941	0.8629
2	0.9509	0.9506	0.9452	0.851
3	0.9504	0.9579	0.9444	0.8655
4	0.9449	0.947	0.9379	0.8634
5	0.9472	0.9467	0.9413	0.8744
6	0.9467	0.9467	0.9457	0.8689
7	0.9532	0.9535	0.9475	0.8377
8	0.9522	0.9532	0.9306	0.8623
9	0.9524	0.9524	0.9478	0.9012
10	0.9441	0.9431	0.9384	0.8672
μ	0.949	0.9497	0.942	0.8654
σ	0.0033	0.0045	0.0053	0.0162

The column names indicate the kernel function corresponding to each SVM and kernel parameters are indicated in brackets. Only the results for the best-performing kernel function (in terms of average AUC across all CV runs) for each set of evaluated parameters are shown. All of the classifiers performed best with a setting of ν = 0.3

ISSN: 1742-4690