Researchers have developed BRIDGE, a new framework for inferring gene regulatory networks from single-cell RNA sequencing data that improves prediction accuracy by refining biological evidence and dynamically gating information transfer.
Motivation and Background
Gene regulatory network (GRN) inference from single-cell RNA sequencing (scRNA-seq) data is critical for uncovering cell-state-specific transcriptional programs. However, as noted in the study published on arXiv, scRNA-seq measurements are "sparse and noisy," and experimentally validated transcription factor (TF)-target interactions remain limited, making reliable inference challenging.
Existing graph neural network (GNN) methods for GRN prediction often rely on biologically unconstrained graph augmentation, such as random edge perturbation. According to the authors, these methods also "insufficiently control information transfer between genes and cells," which can distort regulatory structures and weaken robustness under noisy and weakly supervised settings.
The BRIDGE Framework
To address these issues, the team proposed BRIDGE (Biological Evidence Refinement and Heterogeneous Dynamic Gating for Gene Regulatory Networks). The framework extracts gene and cell representations from the expression matrix and its matrix dual. It then performs contrastive learning in both the gene space and cell space between self and neighbors across a co-expression-refined regulatory view and the original graph.
Critically, BRIDGE applies heterogeneous gated encoding to adaptively regulate information transfer between genes and cells, enabling robust transcription factor-to-target gene prediction. This dynamic gating mechanism is a key differentiator from prior approaches.
Benchmark Performance
Experiments were conducted on benchmark datasets spanning three network types and seven cell types. According to the paper, BRIDGE achieves state-of-the-art AUROC and AUPRC in most settings. A significant result was on Specific networks, where BRIDGE improved average AUPRC by 5% over the second-best baseline, GCLink.
| Metric | BRIDGE vs. GCLink (Specific networks) |
|---|---|
| AUPRC improvement | +5% |
In cross-cell-type few-shot transfer experiments, BRIDGE consistently outperformed both GCLink and GENELink across all six target cell types.
Case Study: Human Embryonic Stem Cells
A case study on human embryonic stem cells (hESC) further supports the biological relevance of BRIDGE's predictions. The authors report that 9 of the top 10 and 46 of the top 100 novel TF-target interactions predicted by BRIDGE were validated by ChIPBase, an independent database of experimentally validated binding events.
Implications for the Research Community
BRIDGE provides a more biologically constrained and robust method for GRN inference. For computational biologists and bioinformatics tool developers, the framework offers an alternative that better preserves regulatory structures while handling noise inherent in scRNA-seq data. The open availability of the method (likely via arXiv and associated code repositories) allows adoption and further benchmarking.
What to Watch
Researchers and practitioners should monitor future releases of BRIDGE with additional benchmarking on larger datasets and integration with downstream analyses such as cell-type classification and drug response prediction.