Skip to main content

Table 4 Median MCC scores of imputation models

From: Analysis of the benefits of imputation models over traditional QSAR models for toxicity prediction

  

Ames

Tox21

Single task

RF

0.520 (0.517–0.523)

0.406 (0.404–0.412)

XGB

0.540 (0.537–0.543)

0.415 (0.412–0.421)

ST-DNN

0.500 (0.495–0.507)

0.415 (0.406–0.419)

Multi-task

XGB-FN

0.677 (0.675–0.682)

0.521 (0.516–0.525)

MT-DNN

0.676 (0.667–0.688)

0.503 (0.493–0.512)

Macau

0.679 (0.677–0.681)

0.385 (0.379–0.388)

  1. Single task models are included as a benchmark. Median scores and interquartile ranges for each technique and dataset on the test set across 20 different random seeds. Before computing the median, the mean across the different assays for a single run was calculated. The best model for each dataset is in bold.