Table 1 Fungus-animal PPI D-SCRIPT benchmark.

From: Using machine learning to predict protein–protein interactions between a zombie ant fungus and its carpenter ant host

Evaluation dataset

AUPR

Precision

Recall

Fly, as published35

0.552

0.798

0.359

Yeast, as published35

0.405

0.706

0.223

Fungus-animal

0.373

0.720

0.169

  1. We evaluated D-SCRIPT on interspecific fungus-animal PPIs using positive interactions from the Host–Pathogen Interaction Database (n = 567) and negative interactions from randomly paired yeast and fly proteins (n = 5670). For this dataset and those published with D-SCRIPT35 with 1 postive:10 negatives, the baseline AUPR is 0.091. Recall is the proportion of true positive examples in the dataset that the model correctly predicted to interact. Precision is the proportion of predicted positives that were indeed correct. Based on these benchmarking data, interspecific PPIs are more difficult for D-SCRIPT to predict, but, we suggest, they still can be detected at rates useful for initial biological investigation and characterization of group-level trends.
Back to article page