Optogenetics enables precise control of neuronal activity using light-sensitive proteins. However, dependence on external light sources introduces key limitations. Bioluminescence-driven approaches offer a promising alternative by generating light internally through proteins such as Aequorin, which emits blue light in response to calcium dynamics. While this emission is compatible with blue light sensitive opsins, its practical effectiveness is limited by the relatively high activation thresholds and functional flexibility of these opsins. Red-sensitive opsins offer a broader range of functionality, including both excitatory and inhibitory control, and certain variants, such as ChRmine, exhibit exceptionally high light sensitivity. These properties make them particularly well suited for low-power bioluminescent systems. We propose a Biological Optical Emitter (BOE) consisting of an Aequorin containing a core surrounded by a polydimethylsiloxane (PDMS) shell embedded with CdSe/ZnS quantum dots to enable wavelength conversion from blue to red. In this design, the native blue emission is not treated as the functional stimulation source; instead, it serves as an internal excitation mechanism for quantum dots, driving red-shifted emission that constitutes the effective output of the system. Our modeling results demonstrate that, despite conversion losses, the system operates effectively as a red-light bioluminescent stimulator, enabling optogenetic activation under power-limited conditions at long distances.
Molecular communication (MC) is a bio-inspired paradigm that models information transfer through molecules and provides an approach to analyze biological processes. Positioning neurological signaling within this framework enables a communication-theoretic view of disorders such as epilepsy. Epileptic seizures originate from microscopic neurotransmitter dynamics. Current prediction methods rely on EEG, which captures only macroscopic activity and often misses early preictal buildup. We propose an MC framework that models microscale synaptic glutamate signaling as a channel shaped by diffusion, clearance, and finite-receptor binding. Our Partial Differential Equations (PDEs) and Chemical Master Equation (CME) combined model yields impulse response and validates stochastic binding. Further analysis shows how synaptic parameters influence detection dynamics. Framing seizure prediction as a threshold-crossing detection problem aligns with latency and reliability metrics in communication theory, positioning MC as a promising framework for next-generation molecular-level seizure detection.
Despite significant efforts to overcome the Debye screening effect, the clinical utility of graphene field-effect transistor (GFET) biosensors remains limited by two key challenges: reliance on externally suspended liquid-gate electrodes that hinder scalable integration and the limited capability of conventional electrical analysis to interpret multiplex biosensing signals in complex biological matrices with satisfactory accuracy. Here, we address both limitations by engineering a printed Ag/AgCl coplanar gate on a PCB-integrated graphene transistor array and coupling the sensing platform with waveform-driven deep learning. Systematic thermal optimization of the printed Ag/AgCl gate establishes 180°C as the optimum annealing condition, producing approximately 50% higher transconductance coupled with significantly reduced device-to-device variability through improved gate-electrolyte coupling and preservation of the electrochemically active Ag/AgCl interface. The optimized transistor array is subsequently functionalized for simultaneous detection of IL-6, IL-8, and TNF-α using AC electrical measurements, mitigating electrolyte screening effects. The platform achieves a practical limit of detection of 1 aM with reproducible responses in serum over a broad analytical range. To overcome concentration overlap, fabrication-induced variability, and serum matrix effects, a multi-task convolutional neural network directly analyzes the complete AC electrical waveform rather than manually extracted electrical features. The proposed framework simultaneously performs cytokine identification and concentration prediction, yielding an overall classification accuracy of 97%, with very high concentration prediction accuracies. To the best of our knowledge, this represents the first demonstration of full-waveform artificial intelligence-assisted multiplex cytokine sensing using PCB-integrated graphene transistor arrays. The convergence of optimized coplanar gate engineering, with waveform-driven intelligence establishes a practical framework for intelligent point-of-care inflammatory diagnostics and next-generation integrated bioelectronic systems.
Receiver localization in mobile molecular communication (MMC) within vessel environments faces significant challenges due to the signal transmission under complex scenarios with anomalous diffusion, fluid flow, and the transmitter in the Internet of Nano Things (IoNT). Traditional localization approaches typically rely on complex mathematical derivations and analytical formulations. This paper proposes a novel deep learning architecture based on the Enformer model, specifically designed for robust and accurate receiver localization in three dimensional scenarios. The proposed architecture utilizes a hybrid of attention and convolutional mechanisms to effectively interpret the received signals sequence which are the molecular absorption time series. A realistic system model is developed, incorporating Poiseuille flow and anomalous diffusion for both molecules and the mobile transmitter, which is then used to generate comprehensive training data. Extensive simulations show that the proposed Enformer-based locator significantly surpasses Poisson maximum-likelihood estimation (MLE) and multi-reference-receiver (MRR) joint-likelihood, deep neural network (DNN) and Transformer models. The model demonstrates faster convergence, maintains superior accuracy and robustness across diverse environmental conditions, such as strong anomalous diffusion, high flow velocity, limited observation time, and the mobility of the transmitter. This research marks a significant step towards enabling precise localization for practical intra-body MMC applications.
The development of high-sensitivity biosensing architectures with integrated signal processing capabilities remains a cornerstone of molecular diagnostics. To address the requirements for ultralow-concentration detection, we developed a programmable and modular dual-amplification DNA fluorescent biosensor (DDFB). This platform features a decoupled architecture that synergistically integrates an upstream entropy-driven catalytic (EDC) circuit for initial signal gain with a downstream hybridization chain reaction (HCR) for high-order reporting. As an enzymefree logic-capable system, the DDFB exhibited an experimentally supported linear range of 100 pM-10 nM for ARG detection, spanning two orders of magnitude, with an estimated LOD of 9.14 pM calculated using the 3σ/S criterion. Furthermore, this detection scheme enables versatile platform expansion through modular design, which exhibits superior performance in the detection of both antibiotic resistance genes (ARGs) and miR-141. Collectively, this research provides a robust, reconfigurable framework for next-generation nanobiosensors and information-processing DNA circuits.
