Overview of the PPIFormer.
A, The source of data, mainly including two aspects: collection from open source databases and literature mining by Europe PMC and PubMed. B, Generating Negative Samples with Double-Dissimilarity Function. C, Pre-trained protein vector are used to encode protein sequences. D, Transformer mechanism is introduced to extract features. E, Matching heuristic was used to capture more information. The schematics in B were created using BioRender (https://biorender.com).