Table 3 Performance of different models in drug-drug interaction prediction

Position-aware deep multi-task learning⁸³

0.526

0.894

multi-class classification 0.7299

DEML⁸⁴

0.52

0.97

0.48

LCM-DS⁸⁷

0.7117

—

—

link prediction as a binary classification task on networks⁸⁸

0.93

0.96

0.82

Similarity-based machine learning support vector machine predictor⁹²

0.68

0.97

—

A probabilistic approach for collective similarity⁹³

0.34

0.96

0.40

DDI-IS-SL⁹⁴

0.9745

—

0.830

DDI-SSL

(Taking DrugBank dataset as an example)⁹⁶

—

0.991 ± 0.002

0.731 ± 0.002

SSI-DDI⁹⁷

0.999

0.9986

0.9237

MSResG⁹⁸

0.798

0.958

0.732

StructNet-DDI⁹⁹

0.9627

0.997

0.944

BRSNMF¹¹⁰

—

—

—

MRMF¹¹¹

0.78

0.97

0.70

ISCMF¹¹²

0.864

0.899

0.885

TMFUF¹¹³

0.526

0.842

—

GRPMF¹¹⁴

0.4975

0.9385

0.9622

Wasserstein Adversarial Autoencoder-based knowledge graph embeddings¹¹⁶

0.5455

0.9527

—

Predicting rich DDI information through graph embedding¹¹⁷

0.3642

—

—

RaGSEs¹¹⁸

0.864

0.899

0.885

SmileGNN¹²⁰

0.9642

0.9995

—

ACDGNN¹²¹

0.9881

0.9835

0.9411

reverse GNN¹²²

0.9328

0.9790

0.8926

SGFNNs¹²³

0.9881

0.9835

0.9411

AutoDDI

(Taking DrugBank dataset as an example)¹²⁴

0.9952 ± 0.0005

0.9953 ± 0.0004

0.9763 ± 0.00004

LLM-DDI¹³⁵

0.7346

0.9571

0.8542

DDI-JUDGE¹³⁶

0.788

0.801

—