Table 1.

Number of true positives in manually evaluated sets

Abstracts selection	Host–pathogen interactions	Phosphorylation-dependent mechanisms
TOP100	99 (99%)	71 (71%)
TOP50	49 (98%)	41 (82%)
TOP25	25 (100%)	19 (76%)
TOP10	10 (100%)	9 (90%)

Manual validations on 200 abstracts. The ranking of two topics was manually validated. The first topic, host–pathogen interactions, was used to rank abstracts related to Arabidopsis thaliana. The second topic, phosphorylation-dependent molecular processes, was used to rank abstracts from three PPI databases (HPRD, MINT and DIP). The proportion of true positives was calculated from the manual validation of the best 100 (TOP100), 50 (TOP50), 25 (TOP25) and 10 (TOP10) abstracts.