Skip to main content
. 2010 Jun 4;38(Web Server issue):W35–W40. doi: 10.1093/nar/gkq415

Table 3.

Validation for detection of specificity sites by SH and mR scored as area under curve (AUC) for the PR plots versus gold-standard specificity sites in the 22 data sets, 7 sets as defined in Table 2 and 15 sets obtained from Chakrabarti and Panchenko (15)

Dataset cbm9 cd00 cd00 cd00 cd00 cd00 cd00 cd00 CN- GPCR GPCR GST IDH/ LacI MDH/ AQP/ nucl rab ras/ ricin serine Smad Aver
120 264 333 363 365 423 985 myc 190 IMDH LDH GLP cycl.a 5/6 ral Wt'd
# positives     7     3     3  12     6  10     4     3  11  21  21     9  14  28     1  23     2  28  12  21     2  29
mR 0.161 0.058 0.006 0.301 0.010 0.055 0.204 0.329 0.037 0.246 0.347 0.156 0.050 0.266 0.063 0.213 0.417 0.540 0.666 0.186 0.078 0.719 0.310
mR ZInline graphic 0.161 0.058 0.006 0.301 0.010 0.055 0.204 0.329 0.037 0.252 0.347 0.156 0.050 0.282 0.063 0.216 0.417 0.539 0.666 0.186 0.078 0.721 0.312
SH. 0.074 0.054 0.003 0.287 0.008 0.119 0.080 0.198 0.067 0.486 0.489 0.242 0.048 0.124 0.125 0.249 0.413 0.602 0.540 0.194 0.261 0.713 0.330
SH ZInline graphic 0.074 0.054 0.003 0.287 0.008 0.119 0.080 0.198 0.067 0.517 0.489 0.242 0.048 0.207 0.125 0.268 0.413 0.602 0.540 0.194 0.261 0.703 0.342
ProteinKeys 0.049 0.008 0.087 0.203 0.010 0.010 0.002 0.034 0.027 0.377 0.505 0.483 0.065 0.301 0.005 0.119 0.011 0.364 0.092 0.276 0.006 0.748 0.287
PROUST-II 0.349 0.079 0.012 0.055 0.011 0.016 0.049 0.058 0.122 0.308  b 0.446 0.089 0.111 0.015 0.187 0.305 0.455 0.378 0.256 0.750 0.723 0.258
SDPpred v.2 0.122 0.126 0.017 0.376 0.012 0.126 0.234 0.509 0.162 0.508 0.508 0.615 0.196 0.146 0.250 0.242 0.413 0.416 0.357 0.201 0.542 0.522 0.333
Xdet 0.352 0.106 0.080 0.366 0.011 0.103 0.196 0.387 0.086 0.125  b 0.117 0.100 0.190 0.033 0.169 0.054 0.350 0.398 0.173 0.105 0.688 0.234
Xdet supc 0.209 0.106 0.019 0.346 0.012 0.189 0.171 0.534 0.101 0.275  b 0.402 0.129 0.207 0.250 0.208 0.292 0.346 0.545 0.193 0.750 0.677 0.279
Average 0.172 0.072 0.026 0.280 0.010 0.088 0.136 0.286 0.078 0.344 0.448 0.318 0.086 0.204 0.103 0.208 0.304 0.468 0.465 0.206 0.314 0.691 0.298

aNucleotidyl cyclase.

bThe GPCR data set is above the maximum of 1000 sequences for these methods.

cSupervised by using subgroupings.

A higher AUC corresponds to better performance. For comparison, predictions by ProteinKeys, PROUST-II, SDPpred v.2 and Xdet are also shown. Best-scoring methods for each data set are in bold. The final column list the average AUCs per method weighted by number of positives, and the bottom row the averages per data set.