Artificial Intelligence (AI) tools such as ChatGPT have transformed our daily life. Usually, these AI tools are based on Neural Networks (NNs) implemented in digital hardware, i.e., not biological or physical NNs like those found in human brains. Instead, NNs are typically implemented as software on top of a Graphics Processing Unit (GPU) or specialized AI hardware. In contrast, this work-in-progress utilizes a biological NN, demonstrating the use of Synthetic Biological Intelligence (SBI) by implementing a toy example - controlling an inverted pendulum. Biological NNs can be beneficial compared to GPUs or specialized AI hardware, especially in terms of energy efficiency and capability on noisy or dynamic tasks. Following the encoding, decoding, reward, and punishment nature of the biological NN, this work presents initial results toward controlling the physical system in future research.
Despite advances in medical research and therapy, cancer remains a major global health issue due to its complexity in treatment. Depending on the type and extent of the disease, tumor removal with cancer treatment drugs without harming surrounding healthy tissue is a critical challenge. The targeted drug delivery (TDD) approach has emerged to increase the effectiveness of treatment while reducing side effects. This paper envisions a hybrid navigation mechanism that utilizes chemotaxis and entropy to direct nanoscale medical agents (NMAs) toward cancerous cells within a tumor microenvironment (TME) to release medication and treat them. The NMAs sense the concentration gradient of the biomarkers in their vicinity to localize the tumor(s). For this study, we used hypoxia as our biomarker, given the higher oxygen consumption by cancerous cells. The numerical results show that the effectiveness and reliability of the hybrid navigation strategy are significantly higher than those of the random and chemotaxis navigation mechanisms. Furthermore, analysis of cancerous cell statistics over time demonstrates that the proposed method eliminates tumors faster than the other two strategies.
Molecular communication (MC) enables information exchange via chemical signals, offering distinct advantages in environments where electromagnetic communication is ineffective. In this work, we present the first experimental demonstration of an MC system with a function of integrated sensing and communication (ISAC). Specifically, alcohol molecules are modulated using on-off keying and detected by a chemical sensor that enables both binary decoding and real-time odor source localization. A mobile robotic platform interprets molecular concentration variations to navigate and recover messages. The results validate the feasibility of this dual function system, paving the way for future MC with embedded sensing capabilities.
In environments where airborne molecular communication (AMC) is used, the choice of surface materials can significantly affect system performance by altering the concentration and persistence of molecular signals. To gain first insights into these material-dependent effects, we performed experiments in a small closed chamber with a high surface-to-volume ratio and compared the observed AMC signals when varying the inner surface between borosilicate glass and polyethylene (PE)-based plastic. In contrast to plastic, the glass environment produced a signal with a more distinct pattern and shorter duration, showing the impact of material in the environment on molecular retention. This behavior can be attributed to differences in surface interactions, in particular the adsorption (and potentially partial absorption) and subsequent desorption of molecules on plastic surfaces. Our results underline the need to explicitly consider the properties of surface materials in the development and modeling of AMC systems.
The Cellular Potts Model (CPM) is a lattice-based computational framework for simulating tissue dynamics such as cell sorting, collective cell migration, and morphogenesis. While the model effectively captures essential biological properties, its high computational cost poses a significant challenge when simulating large-scale multicellular systems over extended simulation times. In our previous work, we proposed a boundary-based algorithm that reduces computation time by limiting copy attempts to cell boundaries and performing local energy calculations. In this paper, we further accelerate CPM simulations with a CPU-GPU hybrid boundary-based algorithm: local energy calculations are offloaded to the GPU to leverage its parallel processing capabilities, while the CPU continues to manage cell boundaries and connectivity checks. Benchmark results demonstrate a substantial reduction in computation time compared to the CPU-based boundary-based algorithm, enabling more efficient and scalable simulations of complex multicellular systems.
