ChemSICal-Net: Timing-Controlled Chemical Reaction Network for Successive Interference Cancellation in Molecular Multiple Access

Abstract

MC networks are envisioned to enable synthetic information exchange between nanoscale biological entities. For many algorithm proposals in the MC research field, the question of implementation at nanoscales and in biological environments remains open. Chemical reaction networks (CRNs) provide a natural framework to model computing processes in biological systems, while detailed simulations capture realistic stochastic effects. In this work, we present ChemSICal-Net, a comprehensive CRN simulation model of a chemical receiver implementing successive interference cancellation (SIC) to differentiate messages from multiple transmitters. We present the structure of the SIC algorithm in the form of basic chemical building blocks and incorporate clocked timing control by a chemical oscillator. We propose an adaptive Bayesian optimization (BO) scheme with a Gaussian process surrogate to find appropriate values for the reaction rate constants and the initial concentrations and show that it outperforms baseline methods from related work based on a fair computational cost metric. Then, the performance of the ChemSICal-Net framework is evaluated stochastically across a range of clock speeds and in different configurations focusing on communication system metrics such as detection accuracy and decision time. Our results highlight that the timing via a chemical clock can improve the detection accuracy by a factor of 2 in scenarios with shorter decision times, which underlines how the trade-off between decision time and detection probability can shape CRN design choices. The BO scheme is shown to reliably optimize parameters for different configurations by approximately one order of magnitude compared to the non-optimized case. Our system reveals the need for a multi-scale approach with external BO and stochastic simulation of molecular reaction dynamics for communication-metric-focused system design.

Figures (22)

• Figure 1: Schematic representation of the ChemSICal-Net framework, showing the DBMC channel with multiple TX, and the RX with block-based chemical computing operations and chemical clock for timing. We showcase the multi-scale design loop: The system can be expressed as ODE, which can be used for initial parameter tuning. Then, comprehensive stochastic simulations form the input for our proposed BO scheme.
• Figure 2: Example of a simple CRN computing the sum of the molecule counts of two input species A and B as the molecule count of a third chemical species C.
• Figure 3: DBMC scenario with $M$ point TX at distances $d_1$, $d_2$ ... $d_M$ from a spherical RX.
• Figure 4: Simplified threshold-based SIC algorithm for DBMC, adapted from wietfeldErrorProbabilityOptimization2024c.
• Figure 5: Simplified structural diagram of the gene-regulatory Pentilator, the cell-cycle-based AM switch oscillator, and a mixed mechanism phosphorylation oscillator. Circles represent molecules/genes, arrows represent relationships such as repression and amplification.