Skip to main content
. 2009 May 8;37(Web Server issue):W141–W146. doi: 10.1093/nar/gkp353

Table 2.

ROC areas of various topics

Topic Positives Negatives Medline Ranker MScanner
Virus contamination in Europe 28 24426 0.99977 0.9075
Microarray and protein aggregation 71 24689 0.99795 0.8724
Radiology (10) 53 47772 0.99748 0.9939
Text mining 312 24777 0.99601 0.9560
Phosphorylation-dependent processes 136 24572 0.99421 0.9867
Systems biology and pathway 407 24609 0.98812 0.9671
Microarray and cancer 8327 24592 0.97041 0.9889
AIDSBio (10) 4099 47746 0.94179 0.9910
PG07 (10) 1611 47758 0.90237 0.9754

MedlineRanker was compared to MScanner for various topics by the mean ROC area after 10-fold cross-validations (the two columns on the right). The same numbers of abstracts in the training set (positives) and in the background set (negatives) were used by both methods.