ProtAttn-QuadNet: An attention-based deep learning framework for protein-protein interaction prediction using ProtBERT embeddings

The provided Dataset.zip ZIP archive contains three files:uniprotkb_reviewed_sequence_Embeddings.csv – This file includes ProtBERT-derived sequence embeddings for 573,661 reviewed protein entries obtained from the UniProtKB database.ppi_balanced_labeled.csv – This file contains 249,814 protein–protein pairs, comprising 124,907 interacting pairs (label = 1) and 124,907 non-interacting pairs (label = 0).ppi_oversampled_labeled.csv – This file includes an oversampled dataset of 1,082,662 protein–protein pairs, consisting of 541,331 interacting pairs (label = 1) and 541,331 non-interacting pairs (label = 0).The code.zip file contains all scripts and implementation details associated with the ProtAttn-QuadNet methodology.