Data-driven inference reveals nonreciprocal interactions driving collective cell migration.

Uncovering the interactions underlying emergent collective behavior in active and living matter remains a major challenge. Multicellular migration exemplifies this problem, as collective motion is controlled by complex cell–cell interactions. Here, we introduce a general data-driven theoretical framework to learn the dynamical interaction rules governing a system's collective stochastic dynamics directly from experimental trajectory data. Applying this inference approach to multicellular migration experiments, we reconstruct minimal many-body stochastic equations of motion for a broad range of distinct cell types: non-cancerous epithelial MCF10A, mesenchymal-like fibrosarcoma HT1080, and mesenchymal breast cancer-derived MDA-MB-231. We find that while healthy epithelial cells coordinate collective migration through reciprocal repulsion and velocity alignment, cancerous mesenchymal cells additionally exhibit non-reciprocal interactions that couple contact to self-propulsion: contact-induced deceleration in fibrosarcoma cells and contact-induced acceleration in breast cancer cells. Conceptually, cells can exploit these non-reciprocal interactions to tune the onset and speed of flocking, underscoring their role in collective migration and potentially enhancing cancer cell invasiveness. More broadly, our framework offers a general and principled route for extracting interaction laws of active and living systems directly from data.