Molecular Communication Channel as a Physical Reservoir Computer

Abstract

Molecular Communication (MC) channels are characterized by significant memory and nonlinear dynamics arising from diffusion and receptor kinetics. While these properties are often treated as impairments to reliable data transmission, this work investigates whether they can instead be exploited as computational resources. We frame a canonical point-to-point MC channel, comprising ligand diffusion and reversible ligand–receptor binding at a spherical receiver, as a Physical Reservoir Computer (PRC). Using deterministic mean-field modeling and particle-based spatial stochastic simulations, we evaluate the MC system’s capability for temporal information processing on standard chaotic time-series benchmarks. Performance is characterized using both task-specific Normalized Root Mean Square Error (NRMSE) and the task-independent Information Processing Capacity (IPC). Our results reveal a non-monotonic dependence of computational capability on key biophysical parameters, including receptor kinetic rates, diffusion coefficient, and transmitter–receiver distance, identifying operating regimes where memory and nonlinearity are balanced. These findings demonstrate the potential of the MC channel as a biologically compatible computational substrate and provide initial design guidelines pointing toward MC networks that co-locate sensing, communication, and computation.