SIGMA: SIgnaling via Golgi Morphology Analysis for pathway-state classification

Abstract

Organelle morphology is increasingly recognized as a computationally accessible readout of molecular state, yet whether subcellular architecture can identify which signaling pathway is active, not merely whether a cell is perturbed, remains untested. Here we introduce SIGMA (SIgnaling via Golgi Morphology Analysis), a framework that reframes pathway identification as a morphological classification problem centered on the Golgi apparatus. SIGMA integrates imaging flow cytometry for population-scale Golgi profiling, teacher-student knowledge distillation to transfer fine morphological cues from limited high-resolution images to an efficient lower-resolution classifier, and mechanistic generalization testing with chemically distinct activators. Applied to MRC-5 fibroblasts across six conditions spanning four signaling cascades plus controls, SIGMA achieved 90.02% cell-level accuracy and correctly assigned five of seven structurally unrelated external compounds to their expected pathway class. These results establish Golgi morphology as a multiplexed, machine-readable fingerprint of signaling state.