Bioinformatics & Graph-Based Learning
Graph and transformer architectures applied to biological interaction data. Evidenced by X-GNN, published at IEEE ICECTE 2026, on human protein–protein interaction prediction.
Biological interaction data is a graph before it is anything else, and treating it as one is why this sits apart from the imaging work. Protein–protein interaction prediction is the concrete case: the object of study is a pair of nodes in a network whose edges are themselves uncertain, and the evidence for an interaction is as much in the neighbourhood as in the pair.
X-GNN, published at the 2026 5th International Conference on Electrical, Computer & Telecommunication Engineering (ICECTE) and carried by IEEE, is an explainable ensemble graph–transformer framework for human protein–protein interaction prediction. It puts graph neural networks and transformer attention in the same ensemble, and the explainability is in the title rather than appended to the abstract: a predicted interaction nobody can trace back to the structure supporting it is still a decision to spend laboratory time on.
This direction is narrower than the imaging one and honest about it — one published paper, and the transformer and graph-based model work from a research internship at AMIRL.