Work

Recurrent latent dynamics in spinal circuits enable a digital twin of autonomic state transitions

Year: 2026

Type: article

Abstract: Physiological signals underlying homeostasis, such as blood pressure, are dynamic and nonlinear, shaped by interactions among diverse neurons. During blood pressure transitions, neurons in the spinal intermediolateral nucleus (IML), a sympathetic hub, show heterogeneous monotonic increases or decreases in firing. Such diversity is intrinsic to homeostatic control but hampers accurate decoding of internal state dynamics. Here, we introduce a manifold-based model that addresses this challenge by directly mapping IML population activity onto physiological state transitions. Individual neurons responded independently with increases or decreases in firing, but these patterns alone did not align with the nonlinear trajectory of blood pressure. Focusing instead on recurrent population dynamics, we derived a low-dimensional circular manifold that captured the cyclical temporal evolution of state transitions. Embedding this manifold as a dynamic reservoir in a recurrent neural network reproduced both trial-by-trial variability in neural population activity and nonlinear blood pressure dynamics, establishing a circuit-level framework for autonomic state decoding.

Source: iScience

DOI: 10.1016/j.isci.2026.117971