|
PPI extraction performance on filtered corpora |
||||||||
| Corpus |
Co-occurrence |
RelEx |
||||||
| P |
ΔP |
F |
ΔF |
P |
ΔP |
F |
ΔF |
|
|
|
||||||||
| AIMed |
0.53 |
0.36 |
0.68 |
0.39 |
0.85 |
0.45 |
0.63 |
0.19 |
| BioInfer |
0.53 |
0.40 |
0.70 |
0.47 |
0.78 |
0.39 |
0.57 |
0.16 |
| HPRD50 |
0.64 |
0.26 |
0.78 |
0.23 |
0.93 |
0.17 |
0.76 |
0.07 |
| IEPA |
0.88 |
0.47 |
0.94 |
0.36 |
1.00 |
0.26 |
0.75 |
0.08 |
| LLL |
0.50 |
0.00 |
0.66 |
0.00 |
0.82 |
0.00 |
0.77 |
0.00 |
|
Precision and F-score for the co-occurrence and RelEx methods on the corpora with only entities that participate in an interaction preserved. Recall is not shown as it is unaffected by this modification. The Δ columns show absolute difference to results without filtering. | ||||||||
Pyysalo et al. BMC Bioinformatics 2008 9(Suppl 3):S6 doi:10.1186/1471-2105-9-S3-S6 |
||||||||