Spatial Interaction Research Articles

Interactive design of new media images and architectural space: taking Hangzhou as an example

ABSTRACT Traditional architectural space design often relies on fixed display boards, paper guides, and static displays. Viewers can only passively receive information and lack opportunities to interact with the displayed content, making it difficult to generate deep participation and personalized experiences. This article applied new media images into the interactive design of architectural spaces, transforming and innovating traditional architectural spaces, and providing audiences with a richer and more attractive spatial experience. The navigation service was set as the architectural space design goal, and analyzed using the Leifeng Tower building in Hangzhou as an example. Building Information Modeling (BIM) was used to digitally model the Leifeng Tower building, accurately simulating the building structure, spatial layout, and landscape information. By adding virtual information annotations and navigation routes, tourists were provided with a more diverse and colorful architectural space experience. The experimental results showed that among 400 tourists, the highest number of tourists using augmented reality navigation was 204, and the average satisfaction score of tourists with augmented reality navigation reached 90. Moreover, presenting the guide content in a multimodal form can better enhance the spatial experience of tourists. The use of augmented reality in architectural spatial interaction design can enrich tourists’ perception and experience in the architectural space, enabling them to interact with the building in a more vivid and interactive way.

MPeat2D – a fully coupled mechanical–ecohydrological model of peatland development in two dimensions

Abstract. Higher-dimensional models of peatland development are required to analyse the influence of spatial heterogeneity and complex feedback mechanisms on peatland behaviour. However, the current models exclude the mechanical process that leads to uncertainties in simulating the spatial variability in the water table position, vegetation composition, and peat physical properties. Here, we propose MPeat2D, a peatland development model in two dimensions, which considers mechanical, ecological, and hydrological processes together with the essential feedback from spatial interactions. MPeat2D employs poroelasticity theory that couples fluid flow and solid deformation to model the influence of peat volume changes on peatland ecology and hydrology. To validate the poroelasticity formulation, the comparisons between numerical and analytical solutions of Mandel's problems for two-dimensional test cases are conducted. The application of MPeat2D is illustrated by simulating peatland growth over 5000 years above a flat and impermeable substrate with free-draining boundaries at the edges, using constant and variable climate. In both climatic scenarios, MPeat2D produces lateral variability in the water table depth, which results in the variation in the vegetation composition. Furthermore, the drop in the water table at the margin increases the compaction effect, leading to a higher value of bulk density and a lower value of active porosity and hydraulic conductivity. These spatial variations obtained from MPeat2D are consistent with the field observations, suggesting plausible outputs from the proposed model. By comparing the results of MPeat2D to a one-dimensional model and a two-dimensional model without the mechanical process, we argue that mechanical–ecohydrological feedbacks are important for analysing spatial heterogeneity, shape, carbon accumulation, and resilience of peatlands.

Linearly Arranged Multi‐π‐Stacked Structure for Efficient Through‐Space Charge‐Transfer Emitters

AbstractNoncovalent spatial interaction has become an intriguing and important tool for constructing optoelectronic molecules. In this study, we linearly attached three conjugated units in a multi π‐stacked manner by using just one trident bridge based on indeno[2,1‐b]fluorene. To achieve this structure, we improved the synthetic approach through double C−H activation, significantly simplifying the preparation process. Due to the proximity of the C10, C11, and C12 sites in indeno[2,1‐b]fluorene, we derived two novel donor|acceptor|donor (D|A|D) type molecules, 2DMB and 2DMFB, which exhibited closely packed intramolecular stacking, enabling efficient through‐space charge transfer. This molecular construction is particularly suitable for developing high‐performance thermally activated delayed fluorescence materials. With donor(s) and acceptor(s) constrained and separated within this spatially rigid structure, elevated radiative transition rates, and high photoluminescence quantum yields were achieved. Organic light‐emitting diodes incorporating 2DMB and 2DMFB demonstrated superior efficiency, achieving maximum external quantum efficiencies of 28.6 % and 16.2 %, respectively.

Mapping the breast tumor microenvironment: proximity analysis reveals spatial relationships between macrophage subtypes and metastasis-initiating cancer cells.

Metastasis is responsible for the majority of cancer-related fatalities. We previously identified specific cancer cell populations responsible for metastatic events which are cytokeratin-14 (CK14) and E-cadherin positive in luminal tumors, and E-cadherin and vimentin positive in triple-negative tumors. Since cancer cells evolve within a complex ecosystem comprised of immune cells and stromal cells, we sought to decipher the spatial interactions of these aggressive cancer cell populations within the tumor microenvironment (TME). We used imaging mass cytometry to detect 36 proteins in tumor microarrays containing paired primary and metastatic lesions from luminal or triple-negative breast cancers (TNBC), resulting in a dataset of 1,477,337 annotated cells. Focusing on metastasis-initiating cell populations, we observed close proximity to specific fibroblast and macrophage subtypes, a relationship maintained between primary and metastatic tumors. Notably, high CK14 in luminal cancer cells and high vimentin in TNBC cells correlated with close proximity to specific macrophage subtypes (CD163intCD206intPDL1intHLA-DR+ or PDL1highARG1high). Our in-depth spatial analysis demonstrates that metastasis-initiating cancer cells consistently colocalizes with distinct cell populations within the TME, suggesting a role for these cell-cell interactions in promoting metastasis.

Linearly Arranged Multi-π-Stacked Structure for Efficient Through-Space Charge-Transfer Emitters.

Noncovalent spatial interaction has become an intriguing and important tool for constructing optoelectronic molecules. In this study, we linearly attached three conjugated units in a multi π-stacked manner by using just one trident bridge based on indeno[2,1-b]fluorene. To achieve this structure, we improved the synthetic approach through double C-H activation, significantly simplifying the preparation process. Due to the proximity of the C10, C11, and C12 sites in indeno[2,1-b]fluorene, we derived two novel donor|acceptor|donor (D|A|D) type molecules, 2DMB and 2DMFB, which exhibited closely packed intramolecular stacking, enabling efficient through-space charge transfer. This molecular construction is particularly suitable for developing high-performance thermally activated delayed fluorescence materials. With donor(s) and acceptor(s) constrained and separated within this spatially rigid structure, elevated radiative transition rates, and high photoluminescence quantum yields were achieved. Organic light-emitting diodes incorporating 2DMB and 2DMFB demonstrated superior efficiency, achieving maximum external quantum efficiencies of 28.6 % and 16.2 %, respectively.

Spatial Dynamic Interaction Effects and Formation Mechanisms of Air Pollution in the Central Plains Urban Agglomeration in China

Accurately identifying the dynamic interaction effects and network structure characteristics of air pollution is essential for effective collaborative governance. This study investigates the spatial dynamic interactions of air pollution among 30 cities in the Central Plains Urban Agglomeration using convergent cross mapping. Social network analysis is applied to assess the overall and node characteristics of the spatial interaction network, while key driving factors are analyzed using an exponential random graph model. The findings reveal that air pollution levels in the Central Plains Urban Agglomeration initially increase before they decrease, with heavily polluted cities transitioning from centralized to sporadic distribution. Among the interactions, Heze’s air pollution impact on Kaifeng was the strongest, while Xinxiang’s impact on Changzhi was the weakest. The emission and receiving effects peaked during 2010–2012. The air pollution interactions among cities exhibit significant network characteristics, with block model results indicating that emitting and receiving relationships are primarily concentrated in the bidirectional spillover plate. Natural factors such as temperature and precipitation significantly influence the spatial interaction network. Economic and social factors like economic level and industrial sector proportion also have a significant impact. However, population density does not influence the spatial interaction network. This study contributes to understanding the spatial network of air pollution, thereby enhancing strategies for optimizing regional collaborative governance efforts to address air pollution.

MRSAPose: Multi-level routing sparse attention for multi-person pose estimation

Multi-person Pose Estimation (MPPE) aims to reconstruct human poses by locating and connecting keypoints of individuals in input images. The variability of human poses and the complexity of scenes make MPPE reliant on both local details and global structures, and the absence of either can lead to the generation of deformed poses. With the emergence of Transformer, the performance of MPPE has been significantly improved. However, due to self-attention computing attention scores between each pair of positions, the current Transformer-based MPPE exhibits high quadratic complexity. To address these issues, this paper proposes a novel pose estimation model, MRSAPose. MRSAPose utilizes Multi-level Routing Sparse Attention (MRSA) to dynamically select relevant regions for attention, reducing computational complexity and mitigating the impact of irrelevant regions. Furthermore, MRSAPose constructs a Transformer-CNN Parallel Interaction Block (T-CP block) through MRSA and Recursive Residual Gated Convolution (Res-gnConv), facilitating parallel learning of global and local information. By relying on multi-level routing algorithms and high-order spatial interactions conducted by recursive processing of adjacent features, T-CP block helps MRSAPose effectively alleviates the issues of occlusion and misalignment in pose estimation. On multiple challenging keypoint datasets, MRSAPose outperforms current state-of-the-art algorithms, particularly excelling in crowded and occluded scenes.

An approach to modeling the spatial interaction of point sets on integer regular grids

The paper demonstrates a generalized approach to modeling the interaction of points and point sets on integer regular grids in metric spaces. The modeling process is based on the definition of spatial relations in "Point-Set" and "Set-Set" pairs, which allows to determine the spatial interaction between arbitrary geometric objects, defined as grid-based geometric models (GGM), which represent point sets on regular grids, generalized schemes of spatial interaction are developed in the work. As a result of the work, software was developed for conducting computer experiments on simulation of interaction in metric spaces of arbitrary dimensions and visualization of their results. The developed approach was tested on data sets of different dimensions. Figs.: 10. Refs.: 11 titles.

Regional drivers of industrial decarbonisation: a spatial econometric analysis of 238 EU regions between 2008 and 2020

ABSTRACT The European context of socio-economic integration and physical proximity likely plays an essential role in explaining the decarbonisation outcomes of industrial sectors. However, there is hardly any spatial regional analysis on CO2 emissions drivers in European countries. This study investigates the role of geographical space and regional determinants in industrial decarbonisation by analysing how socio-economic drivers and their interregional relationships impact industrial carbon emissions in European regions. We employ a spatial panel data econometric model to a novel panel dataset comprising 13 years (2008–20) of carbon emissions from hard-to-abate industrial sectors from 238 NUTS-2 regions across 27 European Union countries. Results indicate the presence of endogenous spatial interactions and high-time persistence between CO2-eq emissions in European Union regions. As such, industrial carbon emissions of regions follow similar patterns to their neighbours, supporting the evolutionary economic geography and growth theory assumptions of the spatial interaction of carbon emissions between regions. Furthermore, the use of a spatial econometric model illustrates the negative direct and spillover effects that higher levels of education and regional investment in research and development have on industrial CO2-eq emissions.

Spatial guided image captioning: Guiding attention with object's spatial interaction

AbstractNowadays relational position embedding is widely used in many large multi‐modal models. It begins with relational captioning (a branch of image captioning) and contains two procedures: geometric modelling and prior attention. However, there are some problems that remain unsolved in the conventional procedures. This paper reviews the shortcomings of geometric modelling and prior attention. Then, a new framework called relational guided transformer (RGT) is proposed to verify the authors' conclusion from the origin of relational position embedding—relational captioning. Specifically, RGT has two simple but effective improvements in geometric modelling and prior attention: (1) A machine‐learned geometric modelling strategy called multi‐task geometric modelling (MTG) is used under multi‐task learning, replacing the original hand‐made geometric feature. (2) The effectiveness of multiple kinds of prior attention is discussed and preserved in a better form, which is called spatial guided attention (SGA) to integrate the geometric prior knowledge. Extensive experiments on MSCOCO and Flickr30k have been performed to investigate the effectiveness of each module and prove our argument. The superiority of the model comparing to the authors' baseline has also been proven on the offline evaluation with the “Karpathy” test split of both datasets.

Tensor Decomposition for Spatiotemporal Mobility Pattern Learning with Mobile Phone Data

Detecting urban mobility patterns is crucial for policymakers in urban and transport planning. Mobile phone data have been increasingly deployed to measure the spatiotemporal variations in human mobility. This work applied non-negative Tucker decomposition (NTD) to mobile phone-based origin–destination (O-D) matrices to explore mobility patterns’ latent spatial and temporal relationships in the province of Liège, Belgium. Four [Formula: see text] traffic tensors have been built for one regular weekday, one regular weekend day, one holiday weekday, and one holiday weekend day, respectively. The proposed method inferred spatial clusters and temporal patterns while interpreting the correlation between spatial clusters and temporal patterns through geographical visualization. As a result, we found the similarity of O-D and destination–origin (D-O) patterns and the symmetry for the trips of the temporal patterns with evening peak and morning peaks on the weekday. Moreover, we investigated the attraction of different spatial clusters with two temporal patterns on a regular weekday and validated the reconstructed demand using population counts and commuting matrices. Finally, the differences in spatial and temporal interactions have been addressed in detail.

Do high housing prices inhibit entrepreneurial activity?—Evidence from Yangtze River Delta cities, China

AbstractContrary to the existing literature that suggests the inhibitory or facilitating effect of housing prices on entrepreneurial dynamism, this paper explores the nonlinear effect of housing prices on entrepreneurship and countervailing effect of entrepreneurial dynamism on housing prices, in addition to investigating the spatial interaction between the two. This paper constructs a simultaneous equation model (SEM) to empirically test the interaction and spillover effects between housing prices and entrepreneurial dynamism. The results show that: (1) there is a two‐way interaction between housing prices and entrepreneurial dynamism; (2) the estimated coefficients of housing prices and the quadratic square term of housing prices (14.285 and −0.744, respectively) are consistent and significant as expected, indicating an inverted U‐shaped effect of housing prices on entrepreneurial dynamism; and (3) in geospatial terms, there are positive spatial spillovers and negative spatial interaction effects for both house prices and entrepreneurial dynamism in Yangtze River Delta cities.

PCB Defect Detection Based on Improved Deep Learning Model

Printed circuit boards (PCBs) play a critical role in electronic products. Ensuring these products’ long-term reliability and consistent performance requires effective PCB defect detection. Although existing deep learning models for PCB defect detection are not highly accurate, they often neglect capability considerations. This paper introduces a precise, fast, and lightweight defect detection model, CCG-YOLO, based on an enhanced YOLOv5 model to address this issue. The enhancements in CCG-YOLO can be summarized as follows: (1) Improved Backbone network: The feature extraction ability of the Backbone network is enhanced by introducing a C3HB module, which fosters spatial interaction capabilities. (2) Lightweight feature fusion network: A lightweight convolution structure called Ghost-Shuffle Convolution is incorporated in the feature fusion network, remarkably reducing model parameters while maintaining performance. (3) Efficient residual networking: To enhance model performance further, a CNeB module is introduced based on the ConvNeXt network, which replaces the C3 module in the Neck. CNeB improves model detection accuracy and reduces the number of model parameters. The combination of these enhancements results in impressive performance. CCG-YOLO achieves mean average precision (mAP@0.5) of 99.5% and 88.75% in mAP@0.5:0.95 on the TDD-Net public dataset. Compared with the original YOLOv5s algorithm, CCG-YOLO offers a 4.24% improvement in mAP@0.5:0.95, a 1[Formula: see text]MB reduction in model size, a 0.472[Formula: see text]M decrease in the number of parameters, a 0.6G floating point operation reduction in computational complexity, and a 120 frames per second real-time inference speed. These experimental results underscore that the proposed model excels in accuracy and speed and has a compact size for PCB defect detection. Moreover, CCG-YOLO is easily deployable on low-end devices, making it well-suited for meeting the real-time requirements of industrial defect detection.

Spatial Econometrics in Agriculture: Modelling Spatial Dependencies in Data

Spatial econometrics in agriculture focuses on modelling spatial dependencies in data, recognizing that agricultural outcomes are often influenced by geographic proximity and spatial interactions. This approach acknowledges that agricultural phenomena, such as crop yields, pest outbreaks, and soil quality, can exhibit spatial patterns that traditional econometric models may overlook. By incorporating spatial elements into econometric analysis, researchers can better understand how neighbouring regions or locations influence each other's agricultural outcomes. This is crucial for policymakers and farmers seeking to optimize resource allocation, manage environmental impacts, and enhance productivity in agriculture. Spatial econometrics provides a robust framework to uncover hidden relationships and spatial interactions within agricultural data, thereby supporting informed decision-making and sustainable agricultural practices in a spatially interconnected world.

A longitudinal analysis of local innovation in Italy: How do proximity measures matter?

ABSTRACT This paper investigates the factors influencing local innovation from a longitudinal perspective while assessing geographical, economic and technological proximity. The research hypotheses concern spatial interactions, spillover effects and proximity measures that best fit innovation patterns and territorial interactions in Italy. The estimation strategy is the spatial Durbin panel model. The optimal specification to handle cross-sectional dependence in the data was derived from statistical tests evaluating (i) individual-specific effects, (ii) time-specific effects and (iii) both individual and time effects. The model was estimated using data from 107 Italian provinces over 2010–2019. The results show that both endogenous and exogenous interaction effects drive innovation processes and the underlying spillovers are global. Economic proximity explains local innovation patterns more effectively than geographical contiguity and technological proximity.

Get to know your neighbors with a SNAQ™: A framework for single cell spatial neighborhood analysis in immunohistochemical images.

Analyzing the local microenvironment of tumor cells can provide significant insights into their complex interactions with their cellular surroundings, including immune cells. By quantifying the prevalence and distances of certain immune cells in the vicinity of tumor cells through a neighborhood analysis, patterns may emerge that indicate specific associations between cell populations. Such analyses can reveal important aspects of tumor-immune dynamics, which may inform therapeutic strategies. This method enables an in-depth exploration of spatial interactions among different cell types, which is crucial for research in oncology, immunology, and developmental biology. We introduce an R Markdown script called SNAQ™ (Single-cell Spatial Neighborhood Analysis and Quantification), which conducts a neighborhood analysis on immunofluorescent images without the need for extensive coding knowledge. As a demonstration, SNAQ™ was used to analyze images of pancreatic ductal adenocarcinoma. Samples stained for DAPI, PanCK, CD68, and PD-L1 were segmented and classified using QuPath. The resulting CSV files were exported into RStudio for further analysis and visualization using SNAQ™. Visualizations include plots revealing the cellular composition of neighborhoods around multiple cell types within a customizable radius. Additionally, the analysis includes measuring the distances between cells of certain types relative to others across multiple regions of interest. The R Markdown files that comprise the SNAQ™ algorithm and the input data from this paper are freely available on the web at https://github.com/AryehSilver1/SNAQ.

62 Prognostic biomarker development using deep learning and spatial proteomics in papillary renal cell carcinoma

Abstract Background Papillary renal cell carcinoma (pRCC) is a prevalent but understudied kidney cancer pathology, constituting 15-20% of cases. While many patients achieve curative therapy using nephrectomy, others go on to develop progressive, metastatic disease with poor overall prognosis. In order to improve the current standards of care for pRCC patients, it is critical to identify new methods to identify high-risk patients likely to require more intense and risk-modulated follow-up. Our goals in this project are two-fold: (1) to identify prognostic patient clusters based on spatial architecture of tumor tissue, and (2) to identify a minimal subset of markers to accurately predict the defined clusters for clinical translation. Methods A tissue microarray of 100 pRCC patients was assayed using multiplex immunofluorescence with a 31-antibody panel with immune- and cancer-related proteins. We have developed novel algorithms to identify spatial neighborhoods and features. These features are then used to identify unsupervised clusters of patients. This new approach to classifying patients allows for a more specific and comprehensive description of each patient tumor’s spatial composition and association with diverse clinical outcomes. Next, we developed attention-based deep learning models capable of predicting these spatially-defined clusters directly from the underlying immunofluorescence images and subsequently profiled these for interpretability by retraining based on individual channels. We assessed each model iteratively on a validation set and a held-out test set for final model selection. We focussed on developing an accurate model with the minimal number of channels to enable development of a cost-efficient biomarker assay for the clinic. Results We performed a systematic analysis of the cellular and spatial phenotypes for each patient in our cohort and identified six (6) major spatial clusters. These clusters are based on the spatial neighborhood composition of the tumors (Fig 1A) and describe the composite effects of cell-cell interactions within a patient tumor. We identified one cluster in particular (cluster #6, mainly associated with spatial interaction of CD163+ macrophages) with a significantly worse prognosis than the other clusters. Using our advanced deep learning models, we have been able to achieve excellent validation accuracy in predicting a variety of clinical phenotypes and spatial features, including patient clusters and histological grade. We wished to then reduce the set of required features and began our efforts in predicting tumor vs. normal based on individual immunofluorescence channels and identified 6 channels with >75% individual predictive accuracy (Fig. 1B). Compositing all 6 markers further improves our accuracy to 90%. This success paves the way for exciting future prospects as we work to extend these efforts to identify a minimal clinically-applicable subset of protein markers. Conclusions The combination of computational spatial proteomics and deep learning models has the potential to identify new biomarkers specific to the spatial organization of patient tumors. We have identified a particular patient cluster with multiple associations with CD163+ macrophages demonstrating poor prognosis; selecting these patients may allow for tailored and more informed patient care. We are also working to enhance our deep learning models to circumvent the need for stepwise spatial analysis and, instead, allow for direct-from-image assessment of patients for particular spatial features indicative of various prognostic indicators. Our identified set of 6 markers serves as a basis for further refinements as we work to identify a minimal set of relevant markers able to predict various clinical outcomes from immunofluorescence imaging, offering a promising new avenue in the field of cancer pathology and biomarker development. DOD CDMRP Funding: yes

Modeling tuberculosis transmission flow in China, 2010–2012

BackgroundChina has the third largest number of TB cases in the world, and the average annual floating population in China is more than 200 million, the increasing floating population across regions has a tremendous potential for spreading infectious diseases, however, the role of increasing massive floating population in tuberculosis transmission is yet unclear in China.Methods29,667 tuberculosis flow data were derived from the new smear-positive pulmonary tuberculosis cases in China. Spatial variation of TB transmission was measured by geodetector q-statistic and spatial interaction model was used to model the tuberculosis flow and the regional socioeconomic factors.ResultsTuberculosis transmission flow presented spatial heterogeneity. The Pearl River Delta in southern China and the Yangtze River Delta along China’s east coast presented as the largest destination and concentration areas of tuberculosis inflows. Socioeconomic factors were determinants of tuberculosis flow. Some impact factors showed different spatial associations with tuberculosis transmission flow. A 10% increase in per capita GDP was associated with 10.2% in 2010 or 2.1% in 2012 decrease in tuberculosis outflows from the provinces of origin, and 1.2% in 2010 or 0.5% increase in tuberculosis inflows to the destinations and 18.9% increase in intraprovincial flow in 2012. Per capita net income of rural households and per capita disposable income of urban households were positively associated with tuberculosis flows. A 10% increase in per capita net income corresponded to 14.0% in 2010 or 3.6% in 2012 increase in outflows from the origin, 44.2% in 2010 or 12.8% increase in inflows to the destinations and 47.9% increase in intraprovincial flows in 2012. Tuberculosis incidence had positive impacts on tuberculosis flows. A 10% increase in the number of tuberculosis cases corresponded to 2.2% in 2010 or 1.1% in 2012 increase in tuberculosis inflows to the destinations, 5.2% in 2010 or 2.0% in 2012 increase in outflows from the origins, 11.5% in 2010 or 2.2% in 2012 increase in intraprovincial flows.ConclusionsTuberculosis flows had clear spatial stratified heterogeneity and spatial autocorrelation, regional socio-economic characteristics had diverse and statistically significant effects on tuberculosis flows in the origin and destination, and income factor played an important role among the determinants.

Developing Theoretical Models for Atherosclerotic Lesions: A Methodological Approach Using Interdisciplinary Insights.

Atherosclerosis, a leading cause of cardiovascular disease, necessitates advanced and innovative modeling techniques to better understand and predict plaque dynamics. The present work presents two distinct hypothetical models inspired by different research fields: the logistic map from chaos theory and Markov models from stochastic processes. The logistic map effectively models the nonlinear progression and sudden changes in plaque stability, reflecting the chaotic nature of atherosclerotic events. In contrast, Markov models, including traditional Markov chains, spatial Markov models, and Markov random fields, provide a probabilistic framework to assess plaque stability and transitions. Spatial Markov models, visualized through heatmaps, highlight the spatial distribution of transition probabilities, emphasizing local interactions and dependencies. Markov random fields incorporate complex spatial interactions, inspired by advances in physics and computational biology, but present challenges in parameter estimation and computational complexity. While these hypothetical models offer promising insights, they require rigorous validation with real-world data to confirm their accuracy and applicability. This study underscores the importance of interdisciplinary approaches in developing theoretical models for atherosclerotic plaques.

Physical modeling of embryonic transcriptomes identifies collective modes of gene expression.

Starting from one totipotent cell, complex multicellular organisms form through a series of differentiation and morphogenetic events, culminating in a multitude of cell types arranged in a functional and intricate spatial pattern. To do so, cells coordinate with each other, resulting in dynamics which follow a precise developmental trajectory, constraining the space of possible embryo-to-embryo variation. Using recent single-cell sequencing data of early ascidian embryos, we leverage natural variation together with modeling and inference techniques from statistical physics to investigate development at the level of a complete interconnected embryo - an embryonic transcriptome. After developing a robust and biophysically motivated approach to identifying distinct transcriptomic states or cell types, a statistical analysis reveals correlations within embryos and across cell types demonstrating the presence of collective variation. From these intra-embryo correlations, we infer minimal networks of cell-cell interactions, which reveal the collective modes of gene expression. Our work demonstrates how the existence and nature of spatial interactions along with the collective modes of expression that they give rise to can be inferred from single-cell gene expression measurements, opening up a wider range of biological questions that can be addressed using sequencing-based modalities.