Skip to main content

Table 2 Parameters of the modelling functions for the non-electronic dataset

From: Metrics for GO based protein semantic similarity: a systematic evaluation

  LRBS RRBS
  Mean Stdev Res. Mean Stdev Res.
simGIC 2,4 0,33 0,58 0,25 0,32 0,65
simUI 2,4 0,35 0,45 0,29 0,30 0,50
Resnik's measure Avg 2,3 0,31 0,34 −1,05 0,77 0,46
  Max 2,4 0,34 0,58 0,08 0,41 0,64
  BMA 2,4 0,34 0,54 −0,11 0,58 0,64
  GraSM 2,4 0,30 0,40 −0,53 1,16 0,56
Lin's measure Avg 2,3 0,33 0,32 −1,24 0,78 0,42
  Max 2,4 0,35 0,50 0,18 0,34 0,54
  BMA 2,4 0,35 0,49 0,01 0,51 0,57
  GraSM 2,4 0,31 0,40 0,14 0,60 0,54
Jiang & Conrath's measure Avg 2,3 0,33 0,21 0,25 0,20 0,24
  Max 2,4 0,33 0,29 0,32 0,17 0,29
  BMA 2,4 0,33 0,29 0,32 0,21 0,32
  GraSM 2,4 0,30 0,31 0,39 0,27 0,39
  1. For each semantic similarity measure in the non-electronic dataset, and with each of the sequence similarity metrics (LRBS and RRBS), the mean and standard deviation parameters for the normal cumulative distribution function used to model it are shown, as well as the global resolution of the measure. The variability on the normal parameters with RRBS is evident because the fit is somewhat artificial, and does not reflect the fact that the behaviour of measures is visibly isomorphic. The highest resolutions, corresponding to simGIC and Resnik's measure with the maximum and BMA approaches, are highlighted in bold (Mean: mean of the N CDF ; Stdev: standard deviation of the N CDF ; Res: resolution of each measure; LRBS: log reciprocal BLAST score; RRBS: relative reciprocal BLAST score).