SpaTRACE: Spatiotemporal recurrent auto-encoder for reconstructing signaling and regulatory networks from spatiotemporal transcriptomics data

Cell–cell communication (CCC) drives the coordinated cellular dynamics underlying development, regeneration, and disease. Recent advances in spatiotemporal transcriptomics now enable measurement of gene expression within both spatial context and developmental progression. However, most existing CCC methods depend heavily on curated ligand–receptor (LR) databases and assume steady-state expression, limiting their applicability to understudied species and preventing robust inference of dynamic signaling cascades. Here, we introduce SpaTRACE, a transformer-based Granger-style recurrent autoencoder for de novo reconstruction of signaling and gene-regulatory interactions directly from developmental spatiotemporal transcriptomics data. SpaTRACE jointly embeds LR pairs, transcription factors (TFs), and target genes (TGs) into a unified latent space that integrates spatial neighborhoods with temporal causality. Trained to predict future TG expression along pseudotime-sampled trajectories, the model learns biologically meaningful embeddings that capture LR→TG signaling and TF→TG regulatory influence without relying on pathway priors. Across synthetic datasets, SpaTRACE accurately reconstructs LR–TG interactions, TF–TG regulation, and correct LR pairings, outperforming existing CCC tools—especially under pathway-agnostic settings. Applied to mouse midbrain development and axolotl brain regeneration, SpaTRACE recovers canonical signaling modules, identifies stage-specific transitions, and uncovers previously under-characterized interactions, producing CCC signals with high co-expression fractions and bivariate Moran’s I comparable to leading spatial methods. Together, SpaTRACE establishes a general, statistically powerful framework for dissecting dynamic intercellular communication and regulatory networks from spatiotemporal transcriptomics. A user-friendly implementation is available at: https://github.com/VariaanZhou/SpaTRACE.git .