Skip to main content

Table 4 Classification results on the test sets. Accuracy, specificity and sensitivity are defined in Methods. See [25] for a description of the ROC area.

From: HIV-1 coreceptor usage prediction without multiple alignments: an application of string kernels

Coreceptor usage SVM parameter C Kernel parameter Support vectors Accuracy Specificity Sensitivity ROC area
Blended spectrum kernel        
CCR5 0.04 3 204 96.63% 85.33% 98.75% 98.68%
CXCR4 0.7 9 392 93.68% 96.00% 87.68% 96.59%
CCR5 and CXCR4 2 15 430 94.38% 98.16% 67.05% 90.16%
Local alignment kernel        
CCR5 9 1 200 96.42% 87.55% 98.08% 98.12%
CXCR4 0.02 0.05 321 92.21% 97.56% 78.39% 95.11%
CCR5 and CXCR4 0.5 0.1 399 92.28% 97.20% 56.64% 87.49%
Distant segments kernel        
CCR5 0.4 30 533 96.35% 83.55% 98.75% 98.95%
CXCR4 0.0001 30 577 94.80% 97.56% 87.68% 96.25%
CCR5 and CXCR4 0.2 35 698 95.15% 99.20% 65.89% 90.97%
Perfect deterministic classifier        
CCR5 - - - 99.15% 99.55% 99.08% -
CXCR4 - - - 98.66% 99.70% 95.97% -
CCR5 and CXCR4 - - - 97.96% 99.68% 85.54% -
Distant segments kernel trained on test set        
CCR5 0.3 40 425 98.45% 92.88% 99.5% 99.17%
CXCR4 0.0001 35 611 98.66% 99.70% 95.97% 98.29%
CCR5 and CXCR4 0.0001 40 618 97.96% 99.68% 85.54% 96.27%