Skip to main content

Table 1 Comparison of ROC n statistics for different methods (see Methods for definition).

From: Ensemble approach to predict specificity determinants: benchmarking and validation

Methods

ROC100

ROC200

ROC500

ROC1000

ROC3000

SPEER

0.12 ± 0.009

0.22 ± 0.014

0.40 ± 0.011

0.54 ± 0.009

0.80 ± 0.005

GroupSim

0.08 ± 0.011

0.20 ± 0.013

0.36 ± 0.011

0.53 ± 0.009

0.78 ± 0.005

MultiRELIEF

0.11 ± 0.010

0.16 ± 0.009

0.32 ± 0.014

0.50 ± 0.011

0.78 ± 0.006

SDPpred

0.04 ± 0.005

0.08 ± 0.008

0.22 ± 0.008

0.33 ± 0.005

0.70 ± 0.007

SPEL*

0.02 ± 0.010

0.06 ± 0.008

0.16 ± 0.010

0.30 ± 0.012

0.62 ± 0.005

  1. * SPEL algorithm performs automatic clustering along with the subsite identification.