Unraveling hidden interactions in complex systems with deep learning

Physics-informed neural networks for enhanced reference evapotranspiration estimation in Morocco: Balancing semi-physical models and deep learning

Reference evapotranspiration (ETo) is essential for agricultural water management, crop productivity, and irrigation systems. The Penman-Monteith (PM) equation is the standard method for estimating ETo, but its data-intensive nature makes it impractical, especially in situations where the cost of full standardized weather station is prohibitive, maintenance is inadequate, or data quality and continuity are compromised. To overcome those limitations, various semi-physical (SP) and empirical models with limited weather parameters were developed. In this context, artificial intelligence methods for ETo estimation are gaining more attention, balancing simplicity, minimal data requirements, and high accuracy. However, their data-driven nature raises concerns regarding explainability, trustworthiness, adherence to bio-physical laws, and reliability in operational settings. To address this issue, this paper, inspired by the emerging field of Physics-Informed Neural Networks (PINNs), evaluates the integration of SP models into the loss function during the learning process. The new residual loss combines two losses -the data-driven loss and the loss from SP- through a θ parameter, allowing for a convex combination. In-situ agrometeorological data were collected at four automatic weather stations in Tensift Watershed in Morocco, including air temperature (Ta), solar radiation (Rs), relative humidity (RH), and wind speed (Ws). The study integrates Priestley-Taylor (PT), Makkink (MK), Hargreaves-Samani (HS), and Abtew (AB), under four scenarios of data availability levels: (1) Ta, Rs and RH; (2) Ta and Rs; (3) only Ta; and (4) only Rs. The investigation begins with quality-controlling the data and studying the driving factors of ETo. Next, the SP models were calibrated using the CMA-ES optimization algorithm. The proposed PINN was trained and evaluated, first, for the equal contribution scenario (θ = 0.5) and then for θ in the interval [0, 1] with a step of 0.2, thus analyzing the impact of θ on the PINN performance. For the equal contribution, the results showed that the integration had improved the PINN performance in all scenarios in terms of the RMSE and R2, surpassing the fully data-driven model (θ = 0) and the baseline model (θ = 1). Additionally, for all θ within the interval [0.2, 0.8], the PINN required less training to reach optimal values. Finally, the optimal θ values were determined for each scenario using CMA-ES and were 0.258, 0.771, 0.7226 and 0.169 for PT, MK, HS and AB, respectively. While PINNs demonstrated a promising approach for accurate ETo estimation and consequently improved water resource management, the study also represents a step towards implementing controlled, trustworthy, and physics-informed AI in environmental science.

Collision-Avoiding Flocking With Multiple Fixed-Wing UAVs in Obstacle-Cluttered Environments: A Task-Specific Curriculum- Based MADRL Approach

Multiple unmanned aerial vehicles (UAVs) are able to efficiently accomplish a variety of tasks in complex scenarios. However, developing a collision-avoiding flocking policy for multiple fixed-wing UAVs is still challenging, especially in obstacle-cluttered environments. In this article, we propose a novel curriculum-based multiagent deep reinforcement learning (MADRL) approach called task-specific curriculum-based MADRL (TSCAL) to learn the decentralized flocking with obstacle avoidance policy for multiple fixed-wing UAVs. The core idea is to decompose the collision-avoiding flocking task into multiple subtasks and progressively increase the number of subtasks to be solved in a staged manner. Meanwhile, TSCAL iteratively alternates between the procedures of online learning and offline transfer. For online learning, we propose a hierarchical recurrent attention multiagent actor-critic (HRAMA) algorithm to learn the policies for the corresponding subtask(s) in each learning stage. For offline transfer, we develop two transfer mechanisms, i.e., model reload and buffer reuse, to transfer knowledge between two neighboring stages. A series of numerical simulations demonstrate the significant advantages of TSCAL in terms of policy optimality, sample efficiency, and learning stability. Finally, the high-fidelity hardware-in-the-loop (HITL) simulation is conducted to verify the adaptability of TSCAL. A video about the numerical and HITL simulations is available at https://youtu.be/R9yLJNYRIqY.

Network representation of multicellular activity in pancreatic islets: Technical considerations for functional connectivity analysis

Within the islets of Langerhans, beta cells orchestrate synchronized insulin secretion, a pivotal aspect of metabolic homeostasis. Despite the inherent heterogeneity and multimodal activity of individual cells, intercellular coupling acts as a homogenizing force, enabling coordinated responses through the propagation of intercellular waves. Disruptions in this coordination are implicated in irregular insulin secretion, a hallmark of diabetes. Recently, innovative approaches, such as integrating multicellular calcium imaging with network analysis, have emerged for a quantitative assessment of the cellular activity in islets. However, different groups use distinct experimental preparations, microscopic techniques, apply different methods to process the measured signals and use various methods to derive functional connectivity patterns. This makes comparisons between findings and their integration into a bigger picture difficult and has led to disputes in functional connectivity interpretations. To address these issues, we present here a systematic analysis of how different approaches influence the network representation of islet activity. Our findings show that the choice of methods used to construct networks is not crucial, although care is needed when combining data from different islets. Conversely, the conclusions drawn from network analysis can be heavily affected by the pre-processing of the time series, the type of the oscillatory component in the signals, and by the experimental preparation. Our tutorial-like investigation aims to resolve interpretational issues, reconcile conflicting views, advance functional implications, and encourage researchers to adopt connectivity analysis. As we conclude, we outline challenges for future research, emphasizing the broader applicability of our conclusions to other tissues exhibiting complex multicellular dynamics.

Employing graph attention networks to decode psycho-metabolic interactions in Schizophrenia

Schizophrenia is a severe mental disorder characterized by intricate and underexplored interactions between psychological symptoms and metabolic health, presenting challenges in understanding the disease mechanisms and designing effective treatment strategies. To delve deeply into the complex interactions between mental and metabolic health in patients with schizophrenia, this study constructed a psycho-metabolic interaction network and optimized the Graph Attention Network (GAT). This approach reveals complex data patterns that traditional statistical analyses fail to capture. The results show that weight management and medication management play a central role in the interplay between psychiatric disorders and metabolic health. Furthermore, additional analysis revealed significant correlations between the history of psychiatric symptoms and physical health indicators, as well as the key roles of biochemical markers(e.g., triglycerides and low-density lipoprotein cholesterol), which have not been sufficiently emphasized in previous studies. This highlights the importance of medication management approaches, weight management, psychological treatment, and biomarker monitoring in comprehensive treatment and underscores the significance of the biopsychosocial model. This study is the first to utilize a GNN to explore the interactions between schizophrenia symptoms and metabolic features, providing new insights into understanding psychiatric disorders and guiding the development of more comprehensive treatment strategies for schizophrenia.

Technical upgrade of an open-source liquid handler to support bacterial colony screening

The optimization of genetically engineered biological constructs is a key step to deliver high-impact biotechnological applications. The use of high-throughput DNA assembly methods allows the construction of enough genotypic variants to successfully cover the target design space. This, however, entails extra workload for researchers during the screening stage of candidate variants. Despite the existence of commercial colony pickers, their high price excludes small research laboratories and budget-adjusted institutions from accessing such extensive screening capability. In this work we present COPICK, a technical solution to automatize colony picking in an open-source liquid handler Opentrons OT-2. COPICK relies on a mounted camera to capture images of regular Petri dishes and detect microbial colonies for automated screening. COPICK's software can then automatically select the best colonies according to different criteria (size, color and fluorescence) and execute a protocol to pick them for further analysis. Benchmark tests performed for E. coli and P. putida colonies delivers a raw picking performance over pickable colonies of 82% with an accuracy of 73.4% at an estimated rate of 240 colonies/h. These results validate the utility of COPICK, and highlight the importance of ongoing technical improvements in open-source laboratory equipment to support smaller research teams.

Learning physics-consistent particle interactions

Interacting particle systems play a key role in science and engineering. Access to the governing particle interaction law is fundamental for a complete understanding of such systems. However, the inherent system complexity keeps the particle interaction hidden in many cases. Machine learning methods have the potential to learn the behavior of interacting particle systems by combining experiments with data analysis methods. However, most existing algorithms focus on learning the kinetics at the particle level. Learning pairwise interaction, e.g., pairwise force or pairwise potential energy, remains an open challenge. Here, we propose an algorithm that adapts the Graph Networks framework, which contains an edge part to learn the pairwise interaction and a node part to model the dynamics at particle level. Different from existing approaches that use neural networks in both parts, we design a deterministic operator in the node part that allows to precisely infer the pairwise interactions that are consistent with underlying physical laws by only being trained to predict the particle acceleration. We test the proposed methodology on multiple datasets and demonstrate that it achieves superior performance in inferring correctly the pairwise interactions while also being consistent with the underlying physics on all the datasets. While the previously proposed approaches are able to be applied as simulators, they fail to infer physically consistent particle interactions that satisfy Newton's laws. Moreover, the proposed physics-induced graph network for particle interaction also outperforms the other baseline models in terms of generalization ability to larger systems and robustness to significant levels of noise. The developed methodology can support a better understanding and discovery of the underlying particle interaction laws, and hence, guide the design of materials with targeted properties.

A Machine Learning Approach for Predicting the Maximum Spreading Factor of Droplets upon Impact on Surfaces with Various Wettabilities

Drop impact on a dry substrate is ubiquitous in nature and industrial processes, including aircraft de-icing, ink-jet printing, microfluidics, and additive manufacturing. While the maximum spreading factor is crucial for controlling the efficiency of the majority of these processes, there is currently no comprehensive approach for predicting its value. In contrast to the traditional approach based on scaling laws and/or analytical models, this paper proposes a data-driven approach for estimating the maximum spreading factor using supervised machine learning (ML) algorithms such as linear regression, decision tree, random forest, and gradient boosting. For this purpose, a dataset of hundreds of experimental results from the literature and our own—spanning the last thirty years—is collected and analyzed. The dataset was divided into training and testing sets, each representing 70% and 30% of the input data, respectively. Subsequently, machine learning techniques were applied to relate the maximum spreading factor to relevant features such as flow controlling dimensionless numbers and substrate wettability. In the current study, the gradient boosting regression model, capable of handling structured high-dimensional data, is found to be the best-performing model, with an R2-score of more than 95%. Finally, the ML predictions agree well with the experimental data and are valid across a wide range of impact conditions. This work could pave the way for the development of a universal model for controlling droplet impact, enabling the optimization of a wide variety of industrial applications.

Toward modeling metabolic state from single-cell transcriptomics

BACKGROUND

Single-cell metabolic studies bring new insights into cellular function, which can often not be captured on other omics layers. Metabolic information has wide applicability, such as for the study of cellular heterogeneity or for the understanding of drug mechanisms and biomarker development. However, metabolic measurements on single-cell level are limited by insufficient scalability and sensitivity, as well as resource intensiveness, and are currently not possible in parallel with measuring transcript state, commonly used to identify cell types. Nevertheless, because omics layers are strongly intertwined, it is possible to make metabolic predictions based on measured data of more easily measurable omics layers together with prior metabolic network knowledge.

SCOPE OF REVIEW

We summarize the current state of single-cell metabolic measurement and modeling approaches, motivating the use of computational techniques. We review three main classes of computational methods used for prediction of single-cell metabolism: pathway-level analysis, constraint-based modeling, and kinetic modeling. We describe the unique challenges arising when transitioning from bulk to single-cell modeling. Finally, we propose potential model extensions and computational methods that could be leveraged to achieve these goals.

MAJOR CONCLUSIONS

Single-cell metabolic modeling is a rising field that provides a new perspective for understanding cellular functions. The presented modeling approaches vary in terms of input requirements and assumptions, scalability, modeled metabolic layers, and newly gained insights. We believe that the use of prior metabolic knowledge will lead to more robust predictions and will pave the way for mechanistic and interpretable machine-learning models.

Autopoiesis and Its Efficacy—A Metacybernetic View

This paper seeks to explain the nature of autopoiesis and its capacity to be efficacious, and to do this, it uses agency theory as embedded in metacybernetics. Agency, as a generalised intelligent adaptive living system, can anticipate the future once it has internalised a representation of an active contextual situation through autopoiesis. The role of observation and the nature of internalisation will be discussed, explaining that the latter has two states that determine agency properties of cognition. These are assimilation and accommodation. Assimilation is an information process and results in implicit cognition and recognition, whereas accommodation uses assimilated information delivering explicit cognition, recognition, and conscious awareness with rationality. Similarly, anticipation, a required property of the living, has two states, weak and strong, and these correspond to the two states of internalisation. Autopoiesis has various properties identifiable through the lenses of three autonomous but configurable schemas: General Collective Intelligence (GCI), Eigenform, and Extreme Physical Information (EPI). GCI is a pragmatic evolutionary approach concerned with a contextually connected purposeful and relatable set of task processes, each undertaken by a team of subagencies seeking collective fitness. Eigenform is a symbolic approach that is concerned with how observations can be suitably internalised and thus be used as a token to determine future behaviour, and how that which has been internalised can be adopted to anticipate the future. Extreme Physical Information (EPI) is an empirical approach concerned with acquiring information through observation of an unknown parameter through sampling regimes. The paper represents the conceptualisations of each schema in terms of autopoietic efficacy, and explores their configurative possibilities. It will adopt the ideas delivered to enhance explanations of the nature of autopoiesis and its efficacy within metacybernetics, providing a shift in thinking about autopoiesis and self-organisation.