High-quality Models Research Articles

Benchmarking and building DNA binding affinity models using allele-specific and allele-agnostic transcription factor binding data

BackgroundTranscription factors (TFs) bind to DNA in a highly sequence-specific manner. This specificity manifests itself in vivo as differences in TF occupancy between the two alleles at heterozygous loci. Genome-scale assays such as ChIP-seq currently are limited in their power to detect allele-specific binding (ASB) both in terms of read coverage and representation of individual variants in the cell lines used. This makes prediction of allelic differences in TF binding from sequence alone desirable, provided that the reliability of such predictions can be quantitatively assessed.ResultsWe here propose methods for benchmarking sequence-to-affinity models for TF binding in terms of their ability to predict allelic imbalances in ChIP-seq counts. We use a likelihood function based on an over-dispersed binomial distribution to aggregate evidence for allelic preference across the genome without requiring statistical significance for individual variants. This allows us to systematically compare predictive performance when multiple binding models for the same TF are available. To facilitate the de novo inference of high-quality models from paired-end in vivo binding data such as ChIP-seq, ChIP-exo, and CUT&Tag without read mapping or peak calling, we introduce an extensible reimplementation of our biophysically interpretable machine learning framework named PyProBound. Explicitly accounting for assay-specific bias in DNA fragmentation rate when training on ChIP-seq yields improved TF binding models. Moreover, we show how PyProBound can leverage our threshold-free ASB likelihood function to perform de novo motif discovery using allele-specific ChIP-seq counts.ConclusionOur work provides new strategies for predicting the functional impact of non-coding variants.

Current State and Outlook in Medical 3D Printing and the Role of Radiology.

Medical three-dimensional (3D) printing is playing an increasingly important role in clinical practice. The use of 3D printed models in patient care offers a wide range of possibilities in terms of personalized medicine, training and education of medical professionals, and communication with patients. DICOM files from imaging modalities such as CT and MRI provide the basis for the majority of the 3D models in medicine. The image acquisition, processing, and interpretation of these lies within the responsibility of radiology, which can therefore play a key role in the application and further development of 3D printing.The purpose of this review article is to provide an overview of the principles of 3D printing in medicine and summarize its most important clinical applications. It highlights the role of radiology as central to developing and administering 3D models in everyday clinical practice.This is a narrative review article on medical 3D printing that incorporates expert opinions based on the current literature and practices from our own medical centers.While the use of 3D printing is becoming increasingly established in many medical specialties in Germany and is finding its way into everyday clinical practice, centralized "3D printing labs" are a rarity in Germany but can be found internationally. These labs are usually managed by radiology departments, as radiology is a connecting discipline that - thanks to the imaging technology used to produce data for 3D printing - can play a leading role in the application of medical 3D printing. Copying this approach should be discussed in Germany in order to efficiently use the necessary resources and promote research and development in the future. · 3D printing in medicine is a rapidly growing field.. · Image acquisition and processing provides an important basis for high-quality 3D models.. · Radiology, as the specialist discipline responsible for imaging, has a crucial role to play.. · Radiology should play a leading role in the introduction of 3D printing in everyday clinical practice. . · Streckenbach A, Schubert N, Streckenbach F et al. Current State and Outlook in Medical 3 D Printing and the Role of Radiology. Fortschr Röntgenstr 2024; DOI 10.1055/a-2436-7185.

Machine Learning Accelerated Discovery of Entropy-Stabilized Oxide Catalysts for Catalytic Oxidation.

The catalytic properties of unary to ternary metal oxides were already well experimentally explored, and the left space seems like only high entropy metal oxides (HEOs, element types ≥5). However, the countless element compositions make the trial-and-error method of discovering HEO catalysts impossible. Herein, based on the study of the crystal phase and catalytic performance of the ACr2Ox catalyst system, the strong correlation between the single spinel phase and good catalytic activity of CH4 oxidation was inferred owing to the similar element importance sequences, which were acquired by the corresponding high accuracy machine learning models (cross-validation score >0.7). Furthermore, searching for negative data and choosing the proper training data resulted in high-quality regression models to search for better catalysts. Finally, the screened irregular catalyst Ni0.04Co0.48Zn0.36V0.12Cr2Ox with outstanding sulfur and moisture resistance and long-term stability (>7000 h, T90 = 345 °C) envisions the potential of applying the machine learning method to discover HEOs for target processes.

Pan-Draft: automated reconstruction of species-representative metabolic models from multiple genomes

The accurate reconstruction of genome-scale metabolic models (GEMs) for unculturable species poses challenges due to the incomplete and fragmented genetic information typical of metagenome-assembled genomes (MAGs). While existing tools leverage sequence homology from single genomes, this study introduces pan-Draft, a pan-reactome-based approach exploiting recurrent genetic evidence to determine the solid core structure of species-level GEMs. By comparing MAGs clustered at the species-level, pan-Draft addresses the issues due to the incompleteness and contamination of individual genomes, providing high-quality draft models and an accessory reactions catalog supporting the gapfilling step. This approach will improve our comprehension of metabolic functions of uncultured species.

Streamlining management in thoracic trauma: radiomics- and AI-based assessment of patient risks

BackgroundIn blunt chest trauma, patient management is challenging because clinical guidelines miss tools for risk assessment. No clinical scale reliably measures the severity of cases and the chance of complications.AimThe objective of the study was to optimize the management of patients with blunt chest trauma by creating models prognosticating the transfer to the intensive care unit and in-hospital length of stay (LOS).MethodsThe study cohort consisted of 212 cases. We retrieved information on the cases from the hospital’s trauma registry. After segmenting the lungs with Lung CT Analyzer, we performed volumetric feature extraction with data-characterization algorithms in PyRadiomics.ResultsTo predict whether the patient will require intensive care, we used the three groups of findings: ambulance, admission, and radiomics data. When trained on the ambulance data, the models exhibited a borderline performance. The metrics improved after we retrained the models on a combination of ambulance, laboratory, radiologic, and physical examination data (81.5% vs. 94.4% Sn). Radiomics data were the top-accurate predictors (96.3% Sn). Age, vital signs, anthropometrics, and first aid time were the best-performing features collected by the ambulance service. Laboratory findings, AIS scores for the lower extremity, abdomen, head, and thorax constituted the top-rank predictors received on admission to the hospital. The original first-order kurtosis had the highest predictive value among radiomics data. Top-informative radiomics features were derived from the right hemithorax because the right lung is larger. We constructed regression models that can adequately reflect the in-hospital LOS. When trained on different groups of data, the machine-learning regression models showed similar performance (MAE/ROV≈8%). Anatomic scores for the body parts other than thorax and laboratory markers of hemorrhage had the highest predictive value. Hence, the number of injured body parts correlated with the case severity.ConclusionThe study findings can be used to optimize the management of patients with a chest blunt injury as a specific case of monotrauma. The models we built may help physicians to stratify patients by risk of worsening and overcome the limitations of existing tools for risk assessment. High-quality AI models trained on radiomics data demonstrate superior performance.

Unveiling Metabolic Engineering Strategies by Quantitative Heterologous Pathway Design.

Constructing efficient cell factories requires the rational design of metabolic pathways, yet quantitatively predicting the potential pathway for breaking stoichiometric yield limit in hosts remains challenging. This leaves it uncertain whether the pathway yield of various products can be enhanced to surpass the stoichiometric yield limit and whether common strategies exist. Here, a high-quality cross-species metabolic network model (CSMN) and a quantitative heterologous pathway design algorithm (QHEPath) are developed to address this challenge. Through systematic calculations using CSMN and QHEPath, 12,000 biosynthetic scenarios are evaluated across 300 products and 4 substrates in 5 industrial organisms, revealing that over 70% of product pathway yields can be improved by introducing appropriate heterologous reactions. Thirteen engineering strategies, categorized as carbon-conserving and energy-conserving, are identified, with 5 strategies effective for over 100 products. A user-friendly web server is developed to quantitatively calculate and visualize the product yields and pathways, which successfully predicts biologically plausible strategies validated in literature for multiple products.

Modeling High Energy Molecules and Screening to Find Novel High Energy Candidates.

High energy materials (HEMs) play pivotal roles in diverse military and civil-commercial sectors, leveraging their substantial energy generation. Integrating machine learning (ML) into HEM research can expedite the discovery of high-energy compounds, complementing or replacing traditional experimental approaches. This manuscript presents an application of our in-house Iterative Stochastic Elimination (ISE) algorithm to identify HEMs. ISE is a generic algorithm that produces reasonable solutions for highly complex combinatorial problems. In molecular discovery, ISE focuses on physicochemical properties to distinguish between different classes of molecules. Due to its long track record in discovering novel, highly active biomolecules, we decided to apply ISE to another type of molecular discovery: High-energy materials. Two distinct ISE models, Model A (92 HEMs) and Model B (169 HEMs), integrated non-HEMs for comprehensive analysis. The results showcase significant achievements for both Models A and B. Model A identified 69% of active molecules in Model B, of which 62% had the highest score. Model B identified 80% of active molecules in Model A, with 61% having the highest score among those 80%. Subsequently, Model C was developed, merging all active molecules (261) from Models A and B. Statistical data indicate that Model C is a high-quality model. It was used to screen and score nearly 2 million molecules from the Enamine database. We find 66 molecules with the highest score of 0.89, plus 8 with that score which are active molecules included in the learning set of Model C. From the 66 molecules, 21 (32%) contain at least one nitro group. In conclusion, this study positions the ISE algorithm as a potential tool for discovering novel HEM candidates, offering a promising pathway for efficient and sustainable advancements in high-energy materials research.

Shared Accountability Shaping the Destinies of Individual and Groups of Nonhuman Primates

At the California National Primate Research Center (CNPRC), the preferred housing for rhesus macaques involves maintaining them in complex social groups outdoors, primarily for breeding purposes. This functionally appropriate environment promotes effective coping through the expression of species-typical behaviors and important aspects of species-typical social structure, thus enabling normal animal development, higher reproductive success, and the production of high-quality biological models. Despite the benefits, social housing introduces challenges like trauma from aggressive interactions. These challenges necessitate a delicate balance between tolerating some aggression and preventing repeated targeting of individuals. Therefore, the CNPRC has established a multidisciplinary working group of behavioral management experts, veterinarians, animal care, and researchers that meets regularly to review cases of animals that may need to be removed from their social group. We discuss the criteria and decision-making processes employed to manage and mitigate aggression. We describe the systematic approach to identifying at-risk individuals and the comprehensive evaluation process that guides whether to relocate an animal from their groups or not. Considerations include the welfare of the individual and the group’s social stability. This paper provides case studies demonstrating how the working group applies these criteria and processes in practical scenarios, highlighting the complexities and challenges of such decisions.

VFLGAN: Vertical Federated Learning-based Generative Adversarial Network for Vertically Partitioned Data Publication

In the current artificial intelligence (AI) era, the scale and quality of the dataset play a crucial role in training a high-quality AI model. However, good data is not a free lunch and is always hard to access due to privacy regulations like the General Data Protection Regulation (GDPR). A potential solution is to release a synthetic dataset with a similar distribution to that of the private dataset. Nevertheless, in some scenarios, it has been found that the attributes needed to train an AI model belong to different parties, and they cannot share the raw data for synthetic data publication due to privacy regulations. In PETS 2023, Xue et al. [29] proposed the first generative adversary network-based model, VertiGAN, for vertically partitioned data publication. However, after thoroughly investigating, we found that VertiGAN is less effective in preserving the correlation among the attributes of different parties. This article proposes a Vertical Federated Learning-based Generative Adversarial Network, VFLGAN, for vertically partitioned data publication to address the above issues. Our experimental results show that compared with VertiGAN, VFLGAN significantly improves the quality of synthetic data. Taking the MNIST dataset as an example, the quality of the synthetic dataset generated by VFLGAN is 3.2 times better than that generated by VertiGAN w.r.t. the Frechet Distance. We also designed a more efficient and effective Gaussian mechanism for the proposed VFLGAN to provide the synthetic dataset with a differential privacy guarantee. On the other hand, differential privacy only gives the upper bound of the worst-case privacy guarantee. This article also proposes a practical auditing scheme that applies membership inference attacks to estimate privacy leakage through the synthetic dataset.

MemberShield: A framework for federated learning with membership privacy

Federated Learning (FL) allows multiple data owners to build high-quality deep learning models collaboratively, by sharing only model updates and keeping data on their premises. Even though FL offers privacy-by-design, it is vulnerable to membership inference attacks (MIA), where an adversary tries to determine whether a sample was included in the training data. Existing defenses against MIA cannot offer meaningful privacy protection without significantly hampering the model’s utility and causing a non-negligible training overhead. In this paper we analyze the underlying causes of the differences in the model behavior for member and non-member samples, which arise from model overfitting and facilitate MIAs. Accordingly, we propose MemberShield, a generalization-based defense method for MIAs that consists of: (i) one-time preprocessing of each client’s training data labels that transforms one-hot encoded labels to soft labels and eventually exploits them in local training, and (ii) early stopping the training when the local model’s validation accuracy does not improve on that of the global model for a number of epochs. Extensive empirical evaluations on three widely used datasets and four model architectures demonstrate that MemberShield outperforms state-of-the-art defense methods by delivering substantially better practical privacy protection against all forms of MIAs, while better preserving the target model utility. On top of that, our proposal significantly reduces training time and is straightforward to implement, by just tuning a single hyperparameter.

Towards the accurate modelling of antibody-antigen complexes from sequence using machine learning and information-driven docking.

Antibody-antigen complex modelling is an important step in computational workflows for therapeutic antibody design. While experimentally determined structures of both antibody and the cognate antigen are often not available, recent advances in machine learning-driven protein modelling have enabled accurate prediction of both antibody and antigen structures. Here, we analyse the ability of protein-protein docking tools to use machine learning generated input structures for information-driven docking. In an information-driven scenario, we find that HADDOCK can generate accurate models of antibody-antigen complexes using an ensemble of antibody structures generated by machine learning tools and AlphaFold2 predicted antigen structures. Targeted docking using knowledge of the complementary determining regions on the antibody and some information about the targeted epitope allows the generation of high-quality models of the complex with reduced sampling, resulting in a computationally cheap protocol that outperforms the ZDOCK baseline. The source code of HADDOCK3 is freely available at github.com/haddocking/haddock3. The code to generate and analyse the data is available at github.com/haddocking/ai-antibodies. The full runs, including docking models from all modules of a workflow have been deposited in our lab collection (data.sbgrid.org/labs/32/1139) at the SBGRID data repository.

A Flexible Hierarchical Framework for Implicit 3D Characterization of Bionic Devices.

In practical applications, integrating three-dimensional models of bionic devices with simulation systems can predict their behavior and performance under various operating conditions, providing a basis for subsequent engineering optimization and improvements. This study proposes a framework for characterizing three-dimensional models of objects, focusing on extracting 3D structures and generating high-quality 3D models. The core concept involves obtaining the density output of the model from multiple images to enable adaptive boundary surface detection. The framework employs a hierarchical octree structure to partition the 3D space based on surface and geometric complexity. This approach includes recursive encoding and decoding of the octree structure and surface geometry, ultimately leading to the reconstruction of the 3D model. The framework has been validated through a series of experiments, yielding positive results.

Reconstructing of the geometry of Atg13 and Atg101 molecules while assembling the complex

Aim. Associated subproteins which constitute the ATG1 multi-protein complex in plants and mammals, including ULK1 in humans, are orchestral protein kinase atg-units in resistance to stress stimuli across their different nature. The goals of this endeavour were to characterize the molecular nature of the interaction of ATG13 with ATG101, followed by in silico docking to catch the plausible ensuing integration into a multimeric complex ULK1/ATG1, which initiates the assembly of a PAS-preautophagosomal structure in the first step of autophagy initiation. Methods. Protein structures were modeled by homology using AlphaFold, and molecular dynamics (MD) was performed using GROMACS 5.0 with the Charmm36. Results. By implementing computer modeling methods, a complex of protein kinase atg-units for both ATG13 and ATG101 proteins, reflecting the interaction interface and conformational properties, was constructed for detailed interpretation while forming the forthcoming assembly of the ULK1/ATG1 multi-protein platform. Conclusions. This study provides a high-quality model platform for further sequential studies of protein-protein docking and protein-protein interactions with the possibility of reconstructing a model of the full ULK1/ATG1 complex to identify ATG8 binding sites.

Exploring the metabolic profile of A. baumannii for antimicrobial development using genome-scale modeling.

With the emergence of multidrug-resistant bacteria, the World Health Organization published a catalog of microorganisms urgently needing new antibiotics, with the carbapenem-resistant Acinetobacter baumannii designated as "critical". Such isolates, frequently detected in healthcare settings, pose a global pandemic threat. One way to facilitate a systemic view of bacterial metabolism and allow the development of new therapeutics is to apply constraint-based modeling. Here, we developed a versatile workflow to build high-quality and simulation-ready genome-scale metabolic models. We applied our workflow to create a metabolic model for A. baumannii and validated its predictive capabilities using experimental nutrient utilization and gene essentiality data. Our analysis showed that our model iACB23LX could recapitulate cellular metabolic phenotypes observed during in vitro experiments, while positive biomass production rates were observed and experimentally validated in various growth media. We further defined a minimal set of compounds that increase A. baumannii's cellular biomass and identified putative essential genes with no human counterparts, offering new candidates for future antimicrobial development. Finally, we assembled and curated the first collection of metabolic reconstructions for distinct A. baumannii strains and analyzed their growth characteristics. The presented models are in a standardized and well-curated format, enhancing their usability for multi-strain network reconstruction.

Application of Interface Reduction Methods to Rotordynamic Casing Models

Abstract As the use of digital twins for the purpose of structural health monitoring increases, so does the demand for high-quality models. In the area of rotating machines, this means that the casing must also be taken into account. These are often geometrically complex assemblies that require the use of finite element models. The models are then reduced in size by established reduction methods like the Craig-Bampton reduction in order to keep the solution time, e.g. for a forced response analysis as low as possible. In case of nonlinear contact forces, e.g. friction in bolted joints, a secondary reduction step has to be applied. Here, three different interface reduction methods are investigated and used for a rotor-casing assembly including frictional damping. The necessary basics and terms are presented. The performance of the methods is evaluated for a linear model on substructure and assembly level and for a nonlinear model on the assembly level. The implementation of the interface reduction within the Harmonic Balance Method is presented and tested for different unbalance excitation cases.

Novel Approach to Protect Red Revolutionary Heritage Based on Artificial Intelligence Algorithm and Image-Processing Technology

The red revolutionary heritage is a valuable part of China’s historical and cultural legacy, with the potential to generate economic benefits through its thoughtful development. However, challenges such as insufficient understanding, lack of comprehensive planning and layout, and limited protection and utilization methods hinder the full realization of the political, cultural, and economic value of red heritage. To address these problems, this paper thoroughly examines the current state of red revolutionary heritage protection and identifies the problems within the preservation process. Moreover, it proposes leveraging advanced artificial intelligence (AI) technology to repair some damaged image data. Specifically, this paper introduces a red revolutionary cultural relic image-restoration model based on a generative adversarial network (GAN). This model was trained using samples of damaged image and utilizes high-quality models to restore these images effectively. The study also integrates real-world revolutionary heritage images for practical application and assesses its effectiveness through questionnaire surveys. The survey results show that AI algorithms and image-processing technologies hold significant potential in the protection of revolutionary heritage.

Biological characterisation and computational conformation dynamics of putative L-glutaminase YLaM identified from Bacillus licheniformis

ABSTRACT Glutaminases are a unique group of hydrolytic enzymes with potential applications in cancer therapy and the food industry. This study focused on the biological characterisation and computational conformation dynamics of YLaM, a potential L-glutaminase from a halothermotolerant Bacillus licheniformis. The research integrated experimental and computational methods to identify the YLaM sequence and predict its structural and functional properties. YLaM was amplified using a polymerase chain reaction, cloned into a pET-22b(+) vector, sequenced, and submitted to BankIt/NCBI. This experimental work was complemented by in silico analysis, including structural modelling via the I-TASSER web server, molecular docking using AutoDockTools 1.5.6, and molecular dynamic (MD) simulations using GROMACS v5.1.4. The study resulted in the development of a high-quality 3D model of YLaM through homology modelling and structural refinement, enabling a detailed exploration of its binding affinity with L-glutamine (L-Gln). By identifying active site residues using the homotetramer structure of human glutaminase, crucial interactions with L-Gln were validated, confirming its catalytic function. Finally, MD simulations confirmed the stability of the complex under physiological conditions. This combined molecular identification and functional simulations provide comprehensive insights into the physicochemical properties of the enzyme, offering valuable information for subsequent assessments, including recombinant production and biological impact evaluations.

Characterization of Neutrophil Functional Responses to SARS-CoV-2 Infection in a Translational Feline Model for COVID-19.

There is a complex interplay between viral infection and host innate immune response regarding disease severity and outcomes. Neutrophil hyperactivation, including excessive release of neutrophil extracellular traps (NETs), is linked to exacerbated disease in acute COVID-19, notably in hospitalized patients. Delineating protective versus detrimental neutrophil responses is essential to developing targeted COVID-19 therapies and relies on high-quality translational animal models. In this study, we utilize a previously established feline model for COVID-19 to investigate neutrophil dysfunction in which experimentally infected cats develop clinical disease that mimics acute COVID-19. Specific pathogen-free cats were inoculated with SARS-CoV-2 (B.1.617.2; Delta variant) (n = 24) or vehicle (n = 6). Plasma, bronchoalveolar lavage fluid, and lung tissues were collected at various time points over 12 days post-inoculation. Systematic and temporal evaluation of the kinetics of neutrophil activation was conducted by measuring markers of activation including myeloperoxidase (MPO), neutrophil elastase (NE), and citrullinated histone H3 (citH3) in SARS-CoV-2-infected cats at 4 and 12 days post-inoculation (dpi) and compared to vehicle-inoculated controls. Cytokine profiling supported elevated innate inflammatory responses with specific upregulation of neutrophil activation and NET formation-related markers, namely IL-8, IL-18, CXCL1, and SDF-1, in infected cats. An increase in MPO-DNA complexes and cell-free dsDNA in infected cats compared to vehicle-inoculated was noted and supported by histopathologic severity in respiratory tissues. Immunofluorescence analyses further supported correlation of NET markers with tissue damage, especially 4 dpi. Differential gene expression analyses indicated an upregulation of genes associated with innate immune and neutrophil activation pathways. Transcripts involved in activation and NETosis pathways were upregulated by 4 dpi and downregulated by 12 dpi, suggesting peak activation of neutrophils and NET-associated markers in the early acute stages of infection. Correlation analyses conducted between NET-specific markers and clinical scores as well as histopathologic scores support association between neutrophil activation and disease severity during SARS-CoV-2 infection in this model. Overall, this study emphasizes the effect of neutrophil activation and NET release in SARS-CoV-2 infection in a feline model, prompting further investigation into therapeutic strategies aimed at mitigating excessive innate inflammatory responses in COVID-19.

Chromosome-scale genome assembly of the tropical abalone (Haliotis asinina)

Abalone (family Haliotidae) are an ecologically and economically significant group of marine gastropods that can be found in tropical and temperate waters. To date, only a few Haliotis genomes are available, all belonging to temperate species. Here, we provide the first chromosome-scale abalone genome assembly and the first reference genome of the tropical abalone Haliotis asinina. The combination of PacBio long-read HiFi sequencing and Dovetail’s Omni-C sequencing allowed the chromosome-level assembly of this genome, while PacBio Isoform sequencing across five tissue types enabled the construction of high-quality gene models. This assembly resulted in 16 pseudo-chromosomes spanning over 1.12 Gb (98.1% of total scaffolds length), N50 of 67.09 Mb, the longest scaffold length of 105.96 Mb, and a BUSCO completeness score of 97.6%. This study identified 25,422 protein-coding genes and 61,149 transcripts. In an era of climate change and ocean warming, this genome of a heat-tolerant species can be used for comparative genomics with a focus on thermal resistance. This high-quality reference genome of H. asinina is a valuable resource for aquaculture, fisheries, and ecological studies.

Spatiotemporal Differentiation of Carbon Emissions from Logistics Industry at Provincial Scale in China Under the Background of High-quality Economic Development

Since 2010, the Chinese economy has transitioned from a high-speed growth model to a high-quality development model. During this period, the logistics industry has witnessed rapid growth, leading to significant carbon emissions and posing severe threats to the ecological environment. To investigate the spatiotemporal variations in carbon emissions in China's logistics industry, we conducted a correlation analysis using Moran's I index and a bivariate spatial autocorrelation model from 2010 to 2021. Additionally, we employed a geographically and temporally weighted regression model （GTWR） to examine the spatial heterogeneity of factors influencing provincial-level logistics-related carbon emissions. The results indicated that over the study period, there was a shift from insignificant spatial relationships to significant positive spatial correlations among provincial-level logistics carbon emissions in China. Furthermore, varying degrees of spatial clustering were observed. The findings regarding factor heterogeneity revealed that freight turnover volume, per capita GDP of the logistics industry, and infrastructure level exhibited positive spatial correlations with logistics-related carbon emissions, whereas energy intensity showed negative spatial correlations with such emissions. Comparing the results from the geographically weighted regression （GWR） and ordinary least squares regression （OLS）, it was evident that the adjusted R-squared values for the OLS, GWR, and GTWR models were 0.541, 0.567, and 0.838, respectively. This suggests that our adopted GTWR model provided a superior fit and offered better explanations for spatiotemporal heterogeneity between various influencing factors and logistics-related carbon emissions. These research findings can serve as valuable references for formulating province-specific strategies to reduce carbon emissions within China's economy under its high-quality development context.