Abstract 4364135: A Cross-scale Causal Machine Learning Framework Pinpoints <i>Mgl2</i> ⁺ Macrophage Orchestrators of Balanced Arterial Growth

Postnatal development requires precise coordination across molecular, cellular and tissue scales. However, existing methods often fail to resolve the causal hierarchies that underlie this coordination. Transcriptome-wide studies can identify trait-associated genes but cannot predict how gene expression changes propagate through gene–cell–tissue networks. Here, we present CausaLink, a cross-scale causal mapping framework that reconstructs directed networks by integrating transcriptomic and tissue biomechanical data. We applied it to the postnatal pulmonary arterial tissue remodeling, where a surge in arterial flow triggers conversion of low-flow conduits into high-flow conduits. Longitudinal single-cell and bulk transcriptomic profiling of proximal pulmonary artery (P2-P84) mapped dynamic vascular lineage transitions. Our analysis suggested a bifurcation of shared fibroblast–smooth muscle progenitor in the neonatal phase, initiating downstream shifts at both the cellular and tissue levels. In this context, CausaLink identified Mgl2 + macrophages as central regulators of balanced artery growth. Network-based, multi-trait simulation predicted Mgl2 + macrophages promote lumen expansion while preventing pathological wall thinning or thickening. We validated these predictions in a human induced pluripotent stem cell-derived arterial assembloid populated with MGL high macrophages. In this model, MGL high macrophages drove fibroblast expansion and coordinated tissue growth, as predicted. CausaLink enables systemic prediction across biological scales and offers a data-driven approach for therapeutic design in diseases involving disrupted tissue homeostasis.