Skip to main content

Table 6 BioGRID random-slice and time-slice results

From: Neural networks for link prediction in realistic biomedical graphs: a multi-dimensional evaluation of graph embedding-based approaches

   Random slice Time slice
  Node AUC AUC   Avg. Prec AUC AUC   Avg. Prec
Method combination (ROC) (PR) MAP R-prec @ k (ROC) (PR) MAP R-prec @ k
Deep- Average 97.69 97.62 79.24 73.86 99.30 89.40 90.10 68.94 63.30 97.25*
Walk Concat 97.74 97.65 82.48 77.70 99.18 92.12 92.78 71.61 65.96 98.04
LINE Average 98.10* 97.80* 83.13* 78.22* 99.54* 91.86 92.31 72.85 67.76 97.40
  Concat 98.08 97.76 82.94 78.04 99.29 93.55 93.74 73.60 68.57 97.90
node- Average 98.32* 97.97* 85.70* 81.17* 99.38* 95.25 95.43 74.91 70.39 98.26
2vec Concat 98.51 98.26 86.49 81.84 99.49* 93.66 94.66* 73.48 68.77 98.40*
AA N/A 86.10 90.75 70.97 57.65 96.13 77.46 87.69 74.84 61.39 98.10
CN N/A 91.20 94.96 75.72 69.81 99.64 85.07 91.81 76.20 67.73 99.38
JI N/A 90.80 93.95 73.93 68.79 98.59 84.74 90.20 75.60 67.49 97.45
  1. (Bold: best score, *: not statistically different from best)