Table 4 Average text scores (± standard deviation) of DeepSets and DGN variants when using one-hot encodings as protein identifiers
From: Sensitivity analysis on protein-protein interaction networks through deep graph networks
Model | UC | MCC | F1 | ACC | AUROC |
|---|---|---|---|---|---|
DeepSets | 1 | \(.498\scriptstyle \pm .009\) | \(.661\scriptstyle \pm .008\) | \(.778\scriptstyle \pm .007\) | \(.834\scriptstyle \pm .003\) |
DGN | 1 | \(\mathbf {.764\scriptstyle \pm .008}\) | \(\mathbf {.842\scriptstyle \pm .005}\) | \(\mathbf {.896\scriptstyle \pm .004}\) | \(\mathbf {.952\scriptstyle \pm .001}\) |
DeepSets | 2 | \(.427\scriptstyle \pm .032\) | \(.595\scriptstyle \pm .047\) | \(.759\scriptstyle \pm .015\) | \(.794\scriptstyle \pm .026\) |
DGN | 2 | \(\mathbf {.543\scriptstyle \pm .043}\) | \(\mathbf {.690\scriptstyle \pm .039}\) | \(\mathbf {.801\scriptstyle \pm .018}\) | \(\mathbf {.851\scriptstyle \pm .014}\) |
DeepSets | 3 | \(.146\scriptstyle \pm .076\) | \(.284\scriptstyle \pm .107\) | \(.677\scriptstyle \pm .028\) | \(.622\scriptstyle \pm .042\) |
DGN | 3 | \(\mathbf {.253\scriptstyle \pm .058}\) | \(\mathbf {.486\scriptstyle \pm .047}\) | \(\mathbf {.677\scriptstyle \pm .027}\) | \(\mathbf {.651\scriptstyle \pm .049}\) |