Dango: Predicting higher-order genetic interactions

Higher-order genetic interactions have profound implications for understanding the molecular mechanisms of phenotypic variation, yet they remain poorly characterized. Most studies focus on pairwise interactions because high-throughput screens over the vast combinatorial space are challenging. Here, we develop Dango, a computational method based on a self-attention hypergraph neural network, to predict higher-order genetic interactions among groups of genes. As a proof of concept, we provide predictions for over 400 million trigenic interactions in the yeast S. cerevisiae, greatly expanding their quantitative landscape. Dango accurately predicts trigenic interactions and reveals biological functions related to cell growth. We further incorporate protein embeddings and uncertainty estimation to improve biological relevance and interpretability. Moreover, predicted interactions serve as genetic markers for growth responses across diverse conditions. Together, Dango enables a more complete map of complex genetic interactions that shape phenotypic diversity. A record of this paper's transparent peer review process is included in the supplemental information.