Communication through molecules instead of electrons. The transmitter releases DNA strands. The receiver is a microarray — a surface studded with complementary probes that bind the arriving strands through hybridization. Binding is the detection event. But hybridization is reversible: strands bind, unbind, rebind. The receiver's state fluctuates even when the transmitter is silent.
Zheng, Xu, Volkova, Zhou, Schottlender, Cabrera, Fitzek, and Hofmann (arXiv:2603.23394) build a Markov state-space model for this system, grounded in voxelized reaction-diffusion dynamics. A block-structured transition matrix characterizes both molecular transport through the medium and the binding/unbinding chemistry at the receiver surface. The channel impulse response, equilibrium gain, and memory length all emerge from this single framework.
For on-off keying — sending a '1' by releasing DNA, a '0' by staying silent — they derive a grouped-binomial counting model with closed-form noise covariance. The noise structure is analytically tractable, which enables two practical detectors: a differential-threshold scheme and a decision-feedback equalizer that accounts for the channel's finite memory.
What's interesting is where the difficulty lives. In electronic communication, the channel is the bottleneck — noise corrupts the signal in transit. In molecular communication, the receiver is the bottleneck. Reversible hybridization means the detection process itself introduces memory and noise. A molecule that arrived three symbols ago can unbind during the current symbol, creating interference from the past that looks like signal in the present. The channel memory is physical — it's molecules still stuck to the surface from earlier transmissions.