Computational Model Research Articles

Creating cell-specific computational models of stem cell-derived cardiomyocytes using optical experiments.

Human induced pluripotent stem cell-derived cardiomyocytes (iPSC-CMs) have gained traction as a powerful model in cardiac disease and therapeutics research, since iPSCs are self-renewing and can be derived from healthy and diseased patients without invasive surgery. However, current iPSC-CM differentiation methods produce cardiomyocytes with immature, fetal-like electrophysiological phenotypes, and the variety of maturation protocols in the literature results in phenotypic differences between labs. Heterogeneity of iPSC donor genetic backgrounds contributes to additional phenotypic variability. Several mathematical models of iPSC-CM electrophysiology have been developed to help to predict cell responses, but these models individually do not capture the phenotypic variability observed in iPSC-CMs. Here, we tackle these limitations by developing a computational pipeline to calibrate cell preparation-specific iPSC-CM electrophysiological parameters. We used the genetic algorithm (GA), a heuristic parameter calibration method, to tune ion channel parameters in a mathematical model of iPSC-CM physiology. To systematically optimize an experimental protocol that generates sufficient data for parameter calibration, we created in silico datasets by simulating various protocols applied to a population of models with known conductance variations, and then fitted parameters to those datasets. We found that calibrating to voltage and calcium transient data under 3 varied experimental conditions, including electrical pacing combined with ion channel blockade and changing buffer ion concentrations, improved model parameter estimates and model predictions of unseen channel block responses. This observation also held when the fitted data were normalized, suggesting that normalized fluorescence recordings, which are more accessible and higher throughput than patch clamp recordings, could sufficiently inform conductance parameters. Therefore, this computational pipeline can be applied to different iPSC-CM preparations to determine cell line-specific ion channel properties and understand the mechanisms behind variability in perturbation responses.

Leveraging In Silico Structure-Activity Models to Predict Acute Honey Bee (Apis mellifera) Toxicity for Agrochemicals.

In the realm of crop protection products, ensuring the safety of pollinators stands as a pivotal aspect of advancing sustainable solutions. Extensive research has been dedicated to this crucial topic as well as new approach methodologies in toxicity testing. Hence, within the agricultural and chemical industries, prioritizing pollinator safety remains a constant objective during the development of predictive tools. One of these tools includes computational models like quantitative structure-activity relationships (QSARs) that are valuable in predicting the toxicity of chemicals. This research uses bee toxicity data to develop artificial neural network classification models for predicting honey bee acute toxicity. Bee toxicity data from 1542 compounds were used to develop models; the sensitivity and specificity of the best model were 0.90 and 0.91, respectively. These in silico models can aid in the discovery of next-generation crop protection products. These tools can guide the screening and selection of next-generation crop protection molecules with high margins of safety to pollinators, and candidates with favorable sustainability profiles can be identified at the early discovery stage as precursors to in vivo data generation.

A computational model of a thermoelectric device for the correction of cerebral hyperthermia

A computational model of a thermoelectric device for the correction of cerebral hyperthermia

Theoretical insights into rotary mechanism of MotAB in the bacterial flagellar motor

Many bacteria enable locomotion by rotating their flagellum. It has been suggested that this rotation is realized by the rotary motion of the stator unit, MotAB, which is driven by proton transfer across the membrane. Recent cryo-electron microscopy studies have revealed a 5:2 MotAB configuration, in which a MotB dimer is encircled by a ring-shaped MotA pentamer. Although the structure implicates the rotary motion of the MotA wheel around the MotB axle, the molecular mechanisms of rotary motion and how they are coupled with proton transfer across the membrane remain elusive. In this study, we built a structure-based computational model for Campylobacter jejuni MotAB, conducted comprehensive protonation-state-dependent molecular dynamics simulations, and revealed a plausible proton-transfer-coupled rotation pathway. The model assumes rotation-dependent proton transfer, in which proton uptake from the periplasmic side to the conserved aspartic acid in MotB is followed by proton hopping to the MotA proton-carrying site, followed by proton export to the CP. We suggest that, by maintaining two of the proton-carrying sites of MotA in the deprotonated state, the MotA pentamer robustly rotates by ∼36° per proton transfer across the membrane. Our results provide a structure-based mechanistic model of the rotary motion of MotAB in bacterial flagellar motors and provide insights into various ion-driven rotary molecular motors.

Oscillations in an artificial neural network convert competing inputs into a temporal code.

The field of computer vision has long drawn inspiration from neuroscientific studies of the human and non-human primate visual system. The development of convolutional neural networks (CNNs), for example, was informed by the properties of simple and complex cells in early visual cortex. However, the computational relevance of oscillatory dynamics experimentally observed in the visual system are typically not considered in artificial neural networks (ANNs). Computational models of neocortical dynamics, on the other hand, rarely take inspiration from computer vision. Here, we combine methods from computational neuroscience and machine learning to implement multiplexing in a simple ANN using oscillatory dynamics. We first trained the network to classify individually presented letters. Post-training, we added temporal dynamics to the hidden layer, introducing refraction in the hidden units as well as pulsed inhibition mimicking neuronal alpha oscillations. Without these dynamics, the trained network correctly classified individual letters but produced a mixed output when presented with two letters simultaneously, indicating a bottleneck problem. When introducing refraction and oscillatory inhibition, the output nodes corresponding to the two stimuli activate sequentially, ordered along the phase of the inhibitory oscillations. Our model implements the idea that inhibitory oscillations segregate competing inputs in time. The results of our simulations pave the way for applications in deeper network architectures and more complicated machine learning problems.

Computational Modelling and Simulations to Study the Thermal Enhancement in Nanofluid Flow in Undulating Wavy Cavity of a Cylinder: Finite Element Analysis

Computational Modelling and Simulations to Study the Thermal Enhancement in Nanofluid Flow in Undulating Wavy Cavity of a Cylinder: Finite Element Analysis

Minimal model for reservoir computing

A minimal model for reservoir computing is studied. We demonstrate that a reservoir computer exists that emulates given coupled maps by constructing a modularised network. We describe a possible mechanism for collapses of the emulation in the reservoir computing by introducing a measure of finite scale deviation. Such transitory behaviour is caused by either (i) an escape from a finite-time stagnation near an unstable chaotic set, or (ii) a critical transition driven by the effective parameter drift. Our approach reveals the essential mechanism for reservoir computing and provides insights into the design of reservoir computer for practical applications.

Multiscale modelling and characterisation of fused filament fabricated neat and graphene nanoplatelet reinforced G-polymers

AbstractIn this study, the fused filament fabrication of novel biodegradable G-polymer is investigated, including also filaments reinforced by graphene nanoplatelets. A combined framework of computational multiscale modelling and experimental studies is applied to characterise and model the mechanical behaviour of composite and nanocomposite materials, fabricated using a 3D-printer by depositing G-polymer filaments in +45$$^\circ$$ ∘ and $$-45^\circ$$ - 45 ∘ angles. The investigation is performed on both hot-extruded single filaments and 3D-printed specimens. Experimental results of hot-extruded single filaments are utilized on the microscale to extract the macroscale constitutive models. At the microscale level, representative volume elements with different percentages of penetration are analyzed to compute the effective orthotropic elastic material properties. A comprehensive comparison study is conducted using different micromechanical models, multiscale simulation outputs, and the mechanical properties resulting from experiments. The accuracy of the results obtained from the homogenization technique is validated against realistic finite element microstructural models and experimental measurements. The results demonstrate that computational homogenization, when coupled with accurate property characterisation, serves as a reliable tool for predicting the elastic response of 3D-printed parts. Furthermore, the proposed framework reduces the need for extensive experimental replication and lowers manufacturing costs.

From in vitro to in silico: a pipeline for generating virtual tissue simulations from real image data

3D cell culture models replicate tissue complexity and aim to study cellular interactions and responses in a more physiologically relevant environment compared to traditional 2D cultures. However, the spherical structure of these models makes it difficult to extract meaningful data, necessitating advanced techniques for proper analysis. In silico simulations enhance research by predicting cellular behaviors and therapeutic responses, providing a powerful tool to complement experimental approaches. Despite their potential, these simulations often require advanced computational skills and significant resources, which creates a barrier for many researchers. To address these challenges, we developed an accessible pipeline using open-source software to facilitate virtual tissue simulations. Our approach employs the Cellular Potts Model, a versatile framework for simulating cellular behaviors in tissues. The simulations are constructed from real world 3D image stacks of cancer spheroids, ensuring that the virtual models are rooted in experimental data. By introducing a new metric for parameter optimization, we enable the creation of realistic simulations without requiring extensive computational expertise. This pipeline benefits researchers wanting to incorporate computational biology into their methods, even if they do not possess extensive expertise in this area. By reducing the technical barriers associated with advanced computational modeling, our pipeline enables more researchers to utilize these powerful tools. Our approach aims to foster a broader use of in silico methods in disease research, contributing to a deeper understanding of disease biology and the refinement of therapeutic interventions.

A One-Dimensional Computational Model to Identify Operating Conditions and Cathode Flow Channel Dimensions for a Proton Exchange Membrane Fuel Cell

A one-dimensional computational model has been developed that can be used to identify operating conditions for the cathode side of a proton exchange membrane fuel cell such that both the inlet and outlet relative humidity is equal to 100%. By balancing the calculated pressure drop along the cathode side flow channel with the change in molar composition, inlet conditions for the cathode side can be identified with the goal of avoiding channel flooding. The channel length, height, width and the land-to-channel width ratio are input parameters for the model so that it might be used to dimension the cathode flow field. The model can be used to calculate the limiting current density, and we are presenting unprecedented high values as a result of the high pressure drop along the flow channels. Such high current densities can ultimately result in a fuel cell power density beyond the typical value of 1.0–2.0 W/cm2 for automotive fuel cells.

Selectivity analysis of diaminopyrimidine-based inhibitors of MTHFD1, MTHFD2 and MTHFD2L

The mitochondrial enzyme methylenetetrahydrofolate dehydrogenase (MTHFD2) is involved in purine and thymidine synthesis via 1C metabolism. MTHFD2 is exclusively overexpressed in cancer cells but absent in most healthy adult human tissues. However, the two close homologs of MTHFD2 known as MTHFD1 and MTHFD2L are expressed in healthy adult human tissues and share a great structural resemblance to MTHFD2 with 54% and 89% sequence similarity, respectively. It is therefore notably challenging to find selective inhibitors of MTHFD2 due to the structural similarity, in particular protein binding site similarity with MTHFD1 and MTHFD2L. Tricyclic coumarin-based compounds (substrate site binders) and xanthine derivatives (allosteric site binders) are the only selective inhibitors of MTHFD2 reported till date. Nanomolar potent diaminopyrimidine-based inhibitors of MTHFD2 have been reported recently, however, they also demonstrate significant inhibitory activities against MTHFD1 and MTHFD2L. In this study, we have employed extensive computational modeling involving molecular docking and molecular dynamics simulations in order to investigate the binding modes and key interactions of diaminopyrimidine-based inhibitors at the substrate binding sites of MTHFD1, MTHFD2 and MTHFD2L, and compare with the tricyclic coumarin-based selective MTHFD2 inhibitor. The outcomes of our study provide significant insights into desirable and undesirable structural elements for rational structure-based design of new and selective inhibitors of MTHFD2 against cancer.

Unveiling the impact of SUMOylation at K298 site of heat shock factor 1 on glioblastoma malignant progression

BackgroundGlioblastoma (GBM) poses a significant medical challenge due to its aggressive nature and poor prognosis. Mitochondrial unfolded protein response (UPRmt) and the heat shock factor 1 (HSF1) pathway play crucial roles in GBM pathogenesis. Post-translational modifications, such as SUMOylation, regulate the mechanism of action of HSF1 and may influence the progression of GBM. Understanding the interplay between SUMOylation-modified HSF1 and GBM pathophysiology is essential for developing targeted therapies. MethodsWe conducted a comprehensive investigation using cellular, molecular, and in vivo techniques. Cell culture experiments involved establishing stable cell lines, protein extraction, Western blotting, co-immunoprecipitation, and immunofluorescence analysis. Mass spectrometry was utilized for protein interaction studies. Computational modeling techniques were employed for protein structure analysis. Plasmid construction and lentiviral transfection facilitated the manipulation of HSF1 SUMOylation. In vivo studies employed xenograft models for tumor growth assessment. ResultsOur research findings indicate that HSF1 primarily undergoes SUMOylation at the lysine residue K298, enhancing its nuclear translocation, stability, and downstream heat shock protein expression, while having no effect on its trimer conformation. SUMOylated HSF1 promoted the UPRmt pathway, leading to increased GBM cell proliferation, migration, invasion, and reduced apoptosis. In vivo studies have confirmed that SUMOylation of HSF1 enhances its oncogenic effect in promoting tumor growth in GBM xenograft models. ConclusionThis study elucidates the significance of SUMOylation modification of HSF1 in driving GBM progression. Targeting SUMOylated HSF1 may offer a novel therapeutic approach for GBM treatment. Further investigation into the specific molecular mechanisms influenced by SUMOylated HSF1 is warranted for the development of effective targeted therapies to improve outcomes for GBM patients.

Simulating cardiac fluid dynamics in the human heart

Abstract Cardiac fluid dynamics fundamentally involves interactions between complex blood flows and the structural deformations of the muscular heart walls and the thin valve leaflets. There has been longstanding scientific, engineering, and medical interest in creating mathematical models of the heart that capture, explain, and predict these fluid-structure interactions. However, existing computational models that account for interactions among the blood, the actively contracting myocardium, and the valves are limited in their abilities to predict valve performance, capture fine-scale flow features, or use realistic descriptions of tissue biomechanics. Here we introduce and benchmark a comprehensive mathematical model of cardiac fluid-structure interaction (FSI) in the human heart. A unique feature of our model is that it incorporates biomechanically detailed descriptions of all major cardiac structures that are calibrated using tensile tests of human tissue specimens to reflect the heart's microstructure. Further, it is the first FSI model of the heart that provides anatomically and physiologically detailed representations of all four cardiac valves. We demonstrate that this integrative model generates physiologic dynamics, including realistic pressure-volume loops that automatically capture isovolumetric contraction and relaxation, and that its responses to changes in loading conditions are consistent with the Frank-Starling mechanism. These complex relationships emerge intrinsically from interactions within our comprehensive description of cardiac physiology. Such models can serve as tools for predicting the impacts of medical interventions. They also can provide platforms for mechanistic studies of cardiac pathophysiology and dysfunction, including congenital defects, cardiomyopathies, and heart failure, that are difficult or impossible to perform in patients.

Analysis of the aerothermal performance of modern commercial high-pressure turbine rotors using different levels of fidelity

In the field of design and optimization of sophisticated geometries such as film-cooled turbine blades of modern jet engines, Computational Fluid Dynamics (CFD) simulation is commonly employed. As the pursuit for higher turbine entry temperatures intensifies, it becomes crucial for computational analyses to offer accurate predictions of metal temperatures and heat transfer coefficients on these critical components. This study investigates the impact of key physical factors that characterize the operation of these components on the accuracy of the computational model. In particular, the effects of including the exchange of thermal energy between the fluid and solid domains, as well as modelling the unsteady interaction between the rotor and stator are studied. The research focuses on a fully featured one-and-a-half stage high-pressure turbine of a commercial jet engine, utilizing proprietary software for conducting 3D Reynolds-Averaged Navier-Stokes flow simulations. To model the fluid-solid thermal interaction, steady-state Conjugate Heat Transfer (CHT) simulations are performed. The CHT results are then compared with experimental data obtained from a thermal paint test, achieving good levels of agreement. Additionally, phase-lag simulations are executed under adiabatic-wall conditions to evaluate the influence of stator-rotor interaction on near-wall gas temperatures. This work shows that, by modelling the periodic unsteadiness at the stator-rotor interface with a phase-lag technique, the maximum near-wall gas temperature prediction is significantly increased, sometimes more than 100K, with respect to the steady-state model one. Findings from this work also suggest that simplified “strip” source terms used to model the presence of film cooling hole rows on the surface of a blade can be used with satisfactory accuracy only for nominal conditions. The mass flows delivered by each source strip should not be linearly scaled based on mainstream quantities for use in different conditions, but they should be recalculated from scratch for the new conditions.

Increasing the efficiency of coagulation in resonant gaps due to acoustic flow formation

Relevance. The urgent need to eliminate environmental pollution from industrial emissions of various solid particles. At the same time, maximum attention is paid to cleaning exhaust gases from particles of 2.5 microns in size or less. One of the most promising ways to increase the efficiency of existing gas purification equipment in capturing such particles is their coagulation by exposing the gas flow to high-intensity acoustic vibrations of ultrasonic frequency. However, at low concentrations, even at the maximum permissible sound pressure level, the coagulation efficiency of particles smaller than 2.5 microns is insufficient to increase the recovery rate of gas cleaning equipment. Therefore, there is an urgent need to find new ways to further improve the efficiency of ultrasonic coagulation of particles smaller than 2.5 μm. Aim. To determine the conditions for the formation of vortex flows in ultrasonic fields with the maximum ultrasonic influence in terms of sound pressure level. Conducting comparative studies of the coagulation of particles with a size of 2.5 microns with and without vortex flows. This will make it possible to determine the real values of increasing the efficiency of ultrasonic coagulation during turbulization of a gas-dispersed flow by acoustic flows in comparison with coagulation in a uniform ultrasonic field and without it. Objects. Coagulation of particles under the influence of homogeneous and inhomogeneous ultrasonic fields. Methods. Computer modeling of the formed ultrasonic field by the finite element method using harmonic acoustic analysis. The paper considers the experimental method for studying the process of combining particles under the influence of ultrasonic vibrations. To determine the characteristics of an aerosol during experimental studies, a TIPAS-1 meter based on the small-angle scattering method and the spectral transparency method was used. Results. The paper introduces the results of studies of coagulation of particles with a size of 2.5 microns or less in an ultrasonic field formed in resonant gaps by oscillating disk emitters. The authors proposed to increase the efficiency of coagulation in resonant gaps through the use of ultrasonic disk emitters capable of forming alternating zones of maximum and minimum amplitude oscillations in the resonant gaps. The creation of such zones ensured the formation of vortex-type acoustic flows capable of moving particles within the nodal regions of a standing wave and between them. The involvement of small particles in the formed flows made it possible to increase the probability of their collision. It was established that more effective ultrasonic coagulation provides an increase in the degree of inertial capture for particles of 2.5 microns in size by 6% – from 89 to 95%, for particles of 1.5 microns in size by 7% – from 85 to 92%, and for particles of 0.5 microns by 9% – from 76 to 85%.

Enhanced generalization and specialization of brain representations of semantic knowledge in healthy aging

Aging is often associated with a decrease in cognitive capacities. However, semantic memory appears relatively well preserved in healthy aging. Both behavioral and neuroimaging studies support the view that changes in brain networks contribute to this preservation of semantic cognition. However, little is known about the role of healthy aging in the brain representation of semantic categories. Here we used pattern classification analyses and computational models to examine the neural representations of living and non-living word concepts. The results demonstrate that brain representations of animacy in healthy aging exhibit increased similarity across categories, even across different task contexts. This pattern of results aligns with the neural dedifferentiation hypothesis that proposes that aging is associated with decreased specificity in brain activity patterns and less efficient neural resource allocation. However, the loss in neural specificity for different categories was accompanied by increased dissimilarity of item-based conceptual representations within each category. Taken together, the age-related patterns of increased generalization and specialization in the brain representations of semantic knowledge may reflect a compensatory mechanism that enables a more efficient coding scheme characterized by both compression and sparsity, thereby helping to optimize the limited neural resources and maintain semantic processing in the healthy aging brain.

Understanding the effects of mineralization and structure on the mechanical properties of tendon-bone insertion using mesoscale computational modeling

Tendon-bone fibrocartilaginous insertion, or enthesis, is a specialized interfacial region that connects tendon and bone, effectively transferring forces while minimizing stress concentrations. Previous studies have shown that insertion features gradient mineralization and branching fiber structure, which are believed to play critical roles in its excellent function. However, the specific structure-function relationship, particularly the effects of mineralization and structure at the mesoscale fiber level on the properties and function of insertion, remains poorly understood. In this study, we develop mesoscale computational models of the distinct fiber organization at tendon-bone insertions, capturing the branching network from tendon to interface fibers and the different mineralization scales. We specifically analyze three key descriptors: the mineralization scale of interface fibers, the mean, and relative standard deviation of the local branching angles of interface fibers. Tensile test simulations on insertion models with varying mineralization scales of interface fibers and structures are performed to mimic the primary loading condition applied to the insertion. We measure and analyze five representative mechanical properties: Young's modulus, strength, toughness, resilience, and failure strain. Our results reveal that mechanical properties are significantly influenced by the three key descriptors, with tradeoffs observed between mutually exclusive properties. For instance, strength and resilience plateau beyond a certain mineralization scale, while failure strain and Young's modulus exhibit monotonic decreasing and increasing trends, respectively. Consequently, there exists an optimal mineralization scale for toughness due to these tradeoffs. By analyzing the mesoscale deformation and failure mechanisms from simulation trajectories, we identify three fracture regimes closely related to the trends in mechanical properties, supporting the observed tradeoffs. Additionally, we examine in detail the effects of the mean and relative standard deviation of local branching angles on mechanical properties and deformation mechanisms. Overall, our study enhances the fundamental understanding of the composition-structure-function relationships at the tendon-bone insertion, complementing recent experimental studies. The mechanical insights from our work have the potential to guide the future biomimetic design of fibrillar adhesives and interfaces for joining soft and hard materials.

Characterization and modeling of laser transmission welded polyetherketoneketone (PEKK) joints: Influence of process parameters and annealing on weld properties

Welding high-performance thermoplastics has gained popularity across various industries such as automotive, aerospace, and medical. Laser transmission welding (LTW) has emerged as an effective method for joining thermoplastic parts due to its precise control and high joint quality. PAEK (polyaryletherketone) are wide spreading over various industrial applications as a substitute to metals and thermosets when high durability and performance are required. Polyetherketoneketone (PEKK) is one of these PAEK and it has received less attention than PEEK until now. PEKK, being a semi-crystalline thermoplastic, requires additional care during processing due to its propensity to crystallize. This study presents both experimental and numerical investigations into LTW of PEKK molded parts, aiming to understand the influence of welding parameters and crystallinity on weld joint morphology and mechanical properties. PEKK plates, prepared in amorphous and semi-crystalline states, are subjected to LTW using a 975 nm diode laser. Material characterization confirms differences in crystallinity between the samples, which affect their thermal and optical properties, which are crucial for welding. Welding tests are conducted with varying laser power (between 75 and 95 W) and semi-transparent part thickness (2 and 4 mm). The morphology of joints is analysed. Assemblies undergo post-weld annealing treatment to examine its influence on weld crystallinity and consequent mechanical properties. Results reveal an anisotropic distribution of crystallinity within the heat-affected zone (HAZ). The depths of the molten layer (ML) and semi-crystalline layer (scL) vary with laser power and assembly type. A notable decrease in weld strength with laser power is highlighted, while annealing leads to enhanced crystallinity and improved weld strength. Despite variations, high weld strengths are achieved with annealing. Computational modelling elucidates the complex interplay between laser irradiation, temperature distribution, and crystallization kinetics observed experimentally. Overall, this comprehensive investigation provides valuable insights into optimizing LTW parameters for PEKK parts.

A multiscale distributed neural computing model database (NCMD) for neuromorphic architecture

Distributed neuromorphic architecture is a promising technique for on-chip processing of multiple tasks. Deploying the constructed model in a distributed neuromorphic system, however, remains time-consuming and challenging due to considerations such as network topology, connection rules, and compatibility with multiple programming languages. We proposed a multiscale distributed neural computing model database (NCMD), which is a framework designed for ARM-based multi-core hardware. Various neural computing components, including ion channels, synapses, and neurons, are encompassed in NCMD. We demonstrated how NCMD constructs and deploys multi-compartmental detailed neuron models as well as spiking neural networks (SNNs) in BrainS, a distributed multi-ARM neuromorphic system. We demonstrated that the electrodiffusive Pinsky–Rinzel (edPR) model developed by NCMD is well-suited for BrainS. All dynamic properties, such as changes in membrane potential and ion concentrations, can be easily explored. In addition, SNNs constructed by NCMD can achieve an accuracy of 86.67% on the test set of the Iris dataset. The proposed NCMD offers an innovative approach to applying BrainS in neuroscience, cognitive decision-making, and artificial intelligence research.

Artificial intelligence can regulate light and climate systems to reduce energy use in plant factories and support sustainable food production.

Plant factories with artificial lighting (PFALs) can boost food production per unit area but require resources such as carbon dioxide and energy to maintain optimal plant growth conditions. Here we use computational modelling and artificial intelligence (AI) to examine plant-environment interactions across ten diverse global locations with distinct climates. AI reduces energy use by optimizing lighting and climate regulation systems, with energy use in PFALs ranging from 6.42 kWh kg-1 in cooler climates to 7.26 kWh kg-1 in warmer climates, compared to 9.5-10.5 kWh kg-1 in PFALs using existing, non-AI-based technology. Outdoor temperatures between 0 °C and 25 °C favour ventilation-related energy use reduction, with outdoor humidity showing no clear pattern or effect on energy use. Ventilation-related energy savings negatively impact other resource utilization such as carbon dioxide use. AI can substantially enhance energy savings in PFALs and support sustainable food production.