Table 2 Comparison of evaluation performance results using two different evaluation methods

From: Benchmarking heterogeneous network-based methods for drug repurposing

Methods

BNNR

OMC

HGIMC

Datasets

Fdata

Cdata

Ydata

Fdata

Cdata

Ydata

Fdata

Cdata

Ydata

AUC

Li et al. 2024

0.9362

0.9501

0.9511

0.9465

0.9593

0.9503

0.9182

0.9415

0.9499

This study

0.9031

0.9235

0.9168

0.9196

0.9345

0.9143

0.8880

0.9130

0.9096

AUPR

Li et al. 2024

0.9515

0.9629

0.9649

0.9567

0.9680

0.9649

0.9365

0.9541

0.9607

This study

0.2051

0.2254

0.2237

0.2149

0.2321

0.2128

0.1750

0.1966

0.2001

  1. Results of row “Li et al. 2024" are obtained from the evaluation method in8, where the test set consists of a small, balanced subset of positive associations (1s) and an equal number of negative associations (0s). AUC and AUPR of row “This study" are derived from our evaluation approach (detailed in Section 4.8), which considers the entire dataset for the evaluation. The comparison highlights significant differences in performance metrics (AUC and AUPR) for Fdata, Cdata, and Ydata using the top three methods, OMC, BNNR, and HGIMC, under both evaluation strategies.
Back to article page