GRNFomer: Accurate Gene Regulatory Network Inference Using Graph Transformer

Abstract Motivation: Gene Regulatory Networks (GRNs) are crucial for understanding cellular processes, but accurately inferring them from gene expression data remains challenging due to the complex, non-linear interactions between genes and the high dimensionality of the data. We introduce GRN-Former, an advanced graph transformer model designed to accurately infer regulatory relationships between transcription factors and target genes from single-cell RNA-seq transcriptomics data, while supporting generalization across species and cell types. Results: GRNFormer surpasses existing methods in both accuracy and scalability, achieving an AUROC of 90% and an AUPRC of 86% on test datasets. Our case study on human embryonic stem cells highlights its ability to identify biologically meaningful regulatory interactions and pathways. The freely accessible GRNFormer tool streamlines GRN inference, presenting significant potential to drive advancements in omics data analysis and systems biology. Availability: https://github.com/BioinfoMachineLearning/GRNformer.git