BoYueGRN: Zero-shot causal discovery of directed gene regulatory networks from single-cell transcriptomes via amortized inference over synthetic structural causal models
Gene regulatory network (GRN) inference from single-cell RNA-seq conventionally relies on per-dataset optimization. Existing tools must be refit for every new dataset, and the majority fail to infer causal regulatory directions. Here we present BoYueGRN, an amortized causal discovery framework trained exclusively on 10,000 synthetic structural causal models. For any unseen dataset, a single…
Conventional gene regulatory network (GRN) inference from single-cell RNA-seq data requires optimization for each individual dataset. Most existing tools struggle to accurately determine the direction of causal regulatory relationships. In this study, BoYueGRN is introduced as a promising new approach. BoYueGRN uses amortized inference over synthetic structural causal models to train a single framework.
With just one forward pass, BoYueGRN can generate edge probabilities and regulatory directions for any unseen dataset. The model incorporates TF-centric sliding windows with asymmetric fusion to extend its capabilities to the full transcriptome. BoYueGRN shows exceptional zero-shot performance, achieving high directional accuracy on various genomic CRISPRi Perturb-seq screens.
Applying BoYueGRN to cell-type- and stage-specific GRN dynamics across five diseases revealed experimentally testable biological hypotheses, significantly expanding our understanding of disease mechanisms. This framework revolutionizes directed GRN inference by treating it as a train-once, reuse-across-datasets paradigm, making it well-suited for systematic, atlas-scale mapping of regulatory dynamics in human diseases.
Written by urgent.news from bioRxiv's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
