Table 14

The effect on F-score when changing the ground truth of incorrectly annotated pairs with APG and SL kernels
AIMed BioInfer
  Kernel Original Modified Retrained Δm-o Δr-m Original Modified Retrained Δm-o Δr-m
APG (setting A) 56.18 56.61 56.14 0.43 −0.47 60.66 60.87 61.19 0.21 0.32
APG (setting B) 55.29 55.73 56.72 0.44 0.99 60.61 60.83 60.94 0.22 0.11
APG (setting C) 53.20 53.66 53.96 0.46 0.30 59.91 60.36 60.88 0.45 0.52
APG (setting D) 52.30 52.77 52.99 0.47 0.22 59.42 59.90 60.20 0.48 0.30
APG (avg) 54.24 54.69 54.95 0.45 0.26 60.15 60.60 60.80 0.34 0.31
SL 54.48 55.06 55.57 0.58 0.51 59.99 60.46 60.71 0.47 0.25

Modified – using the original model with modified ground truth; retrained – results of a model retrained on the modified ground truth; Δm-o – difference between modified and original; Δr-m – difference between retrained and modified.

Tikk et al.

Tikk et al. BMC Bioinformatics 2013 14:12   doi:10.1186/1471-2105-14-12

Open Data