Table 8.

Summary of Contributions in Studies on Multi-task or Multi-modal Models for Protein-Protein Interactions. Note that each study employed varied datasets, cross-validation methods, and simulation settings for evaluation, making direct comparisons potentially inconclusive. The highest reported accuracy is presented when models were assessed using multiple datasets.

Author	Metrics and Results	Contributions
Capel et al. [101]	AUC: 0.7632 AUPRC: 0.3844	Proposed a multi-task deep learning approach for predicting residues in PPI interfaces.
Li et al. [102]	AUC: 0.895 AUPRC: 0.899	Developed EP-EDL, an ensemble deep learning model for accurate prediction of human essential proteins.
Linder et al. [103]	AUC: 0.96	Introduced scrambler networks to improve the interpretability of neural networks for biological sequences.
Pan et al. [104]	ACC: 0.8947 MCC: 0.7902 AUC: 0.9548	Proposed DWPPI, a network embedding-based approach for PPI prediction in plants.
Peng et al. [105]	AUC: 0.9116 AUPRC: 0.8332	Introduced MTGCN, a multi-task learning method for identifying cancer driver genes.
Schulte-Sasse et al. [106]	AUPRC: 0.76	Developed EMOGI, integrating MULTIOMICS data with PPI networks for cancer gene prediction.
Thi Ngan Dong et al. [107]	AUC: 0.9804 F1: 0.9379	Developed a multitask transfer learning approach for predicting virus-human and bacteria-human PPIs.
Zheng et al. [108]	AUPRC: 0.965	Developed DeepAraPPI, a deep learning framework for predicting PPIs in Arabidopsis thaliana.