Brouard, C., Mourad, R., & Vialaneix, N. (2024). Should we really use graph neural networks for transcriptomic prediction?
is accepted for publication in Briefings in Bioinformatics. In this article, we question the relevance of graph neural networks (GNN) for phenotype predictions from transcriptomic data through a complete controlled benchmark. This replication study emphasizes that simpler (and less demanding models) like multilayer perceptron and random forest are often at pair with GNN for this task. Check out my Publication page to read the article.