Hybrid Artificial-Living Cell Collectives for Wetware Computing

TL;DR

The paper investigates in situ temporal signal processing using a hybrid artificial–living reservoir, coupling programmable artificial cells with a living bacterial population within a diffusive chemical environment. By encoding inputs into local chemical cues and sampling a high-dimensional reservoir state, a linear readout performs Mackey–Glass forecasting, achieving robust short-horizon accuracy (NRMSE ≈ 0.33–0.40 for H = 1–5) and revealing significant intrinsic memory (MC ≈ 32.6). The work demonstrates that non-neural, wetware substrates can serve as programmable reservoirs, with performance governed by fading memory and spatiotemporal dynamics rather than readout capacity alone, motivating future experimental validation and biomedical applications.

Abstract

Living systems continuously sense, integrate, and act on chemical information using multiscale biochemical networks whose dynamics are inherently nonlinear, adaptive, and energy-efficient. Yet, most attempts to harness such "wetware" for external computational tasks have centered on neural tissue and electrical interfaces, leaving the information-processing potential of non-neural collectives comparatively underexplored. In this letter, we study a hybrid artificial-living cell network in which programmable artificial cells write time-varying inputs into a biochemical microenvironment, while a living bacterial collective provides the nonlinear spatiotemporal dynamics required for temporal information processing. Specifically, artificial cells transduce an external input sequence into the controlled secretion of attractant and repellent molecules, thereby modulating the "local biochemical context" that bacteria naturally sense and respond to. The resulting collective bacterial dynamics, together with the evolving molecular fields, form a high-dimensional reservoir state that is sampled coarsely (voxel-wise) and mapped to outputs through a trained linear readout within a physical reservoir computing framework. Using an agent-based in silico model, we evaluate the proposed hybrid reservoir on the Mackey-Glass chaotic time-series prediction benchmark. The system achieves normalized root mean square error (NRMSE) values of approximately 0.33-0.40 for prediction horizons H=1 to 5, and exhibits measurable short-term memory as encoded in the distributed spatiotemporal patterns of bacteria and biochemicals. These results motivate the future exploration of non-neural hybrid cell networks for in situ temporal signal processing towards novel biomedical applications.

Figures (5)

• Figure 1: Architecture of the hybrid artificial-living cell RC implemented in silico via agent-based modeling framework.(a) A scalar input sequence $u[n]$, derived from a chaotic time series, is encoded into time-varying chemical secretion by mobile ACs. (b) These cells function as programmable local transducers within a 3D voxelated environment, where bacteria interact through motility and metabolic dynamics in response to chemical gradients. (c) The reservoir state is captured by sampling voxel-wise molecular concentrations and bacterial density within each processing window. (d) A trained linear readout maps the biological dynamics to a predicted output. (e) Performance is quantified via NRMSE (and correlation) across prediction horizons, assessing how well the reservoir reconstructs the temporal structure of the input process.
• Figure 2: NRMSE heatmap across prediction horizon ($H$) and tapped-delay depth ($k$), evaluated over multiple temporal splits. Lowest error is achieved at short horizons, with gradually increasing error for larger $H$, consistent with fading memory and limited long-range predictability of the hybrid molecular reservoir.
• Figure 3: Median NRMSE versus prediction horizon ($H$) for varying tapped-delay depths ($k$). Temporal embedding ($k>0$) improves prediction, especially at longer horizons, indicating that task-relevant information is distributed across recent reservoir states.
• Figure 4: Median correlation between predicted and target signals versus prediction horizon ($H$). Correlation degrades gradually with $H$, with larger tapped-delay depths maintaining superior long-horizon performance.
• Figure 5: Linking intrinsic memory to forecasting. (Top) Memory curve $R^2(d)$ obtained by reconstructing delayed inputs $u[n-d]$ from the instantaneous reservoir state $\mathbf{r}[n]$, yielding $\mathrm{MC}\approx 32.6$. (Bottom) Median prediction correlation versus horizon $H$ for selected tapped-delay depths. Performance degradation becomes pronounced near $H^\star \approx 0.7\,\mathrm{MC}$, providing an internal consistency check between intrinsic reservoir memory and achievable prediction horizon.