Spatial Deconvolution of Cell Types and Cell States at Scale Utilizing TACIT

Spatial Deconvolution of Cell Types and Cell States at Scale Utilizing TACIT | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Spatial Deconvolution of Cell Types and Cell States at Scale Utilizing TACIT Jinze Liu, Khoa Huynh, Katarzyna Tyc, Bruno Fernandes Matuck, and 17 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-4536158/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 21 Apr, 2025 Read the published version in Nature Communications → Version 1 posted You are reading this latest preprint version Abstract Identifying cell types and states remains a time-consuming, error-prone challenge for spatial biology. While deep learning is increasingly used, it is difficult to generalize due to variability at the level of cells, neighborhoods, and niches in health and disease. To address this, we developed TACIT, an unsupervised algorithm for cell annotation using predefined signatures that operates without training data. TACIT uses unbiased thresholding to distinguish positive cells from background, focusing on relevant markers to identify ambiguous cells in multiomic assays. Using five datasets (5,000,000-cells; 51-cell types) from three niches (brain, intestine, gland), TACIT outperformed existing unsupervised methods in accuracy and scalability. Integrating TACIT-identified cell types with a novel Shiny app revealed new phenotypes in two inflammatory gland diseases. Finally, using combined spatial transcriptomics and proteomics, we discovered under- and overrepresented immune cell types and states in regions of interest, suggesting multimodality is essential for translating spatial biology to clinical applications. Biological sciences/Computational biology and bioinformatics/Computational models Biological sciences/Computational biology and bioinformatics/Statistical methods spatial biology multimodal transcriptomics proteomics artificial intelligence machine learning deep learning multiplex imaging fluorescence microscopy brain intestine salivary gland cell typing single cell analysis spatial multiomics Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Full Text Additional Declarations Yes there is potential Competing Interest. The authors had access to the study data and reviewed and approved the final manuscript. Although the authors view each of these as noncompeting financial interests, KMB, QTE, BFM, and BMW are all active members of the Human Cell Atlas; furthermore, KMB is a scientific advisor at Arcato Laboratories and Orange Grove Bio. All other authors declare no competing interests. Supplementary Files Extended1.pdf Extended2.pdf Extended3.pdf Extended4.pdf Extended5.pdf Extended6.pdf Extended7.pdf ExtendedData8.xlsx nreditorialpolicychecklistflatTACIT.pdf Editorial Policy Checklist Cite Share Download PDF Status: Published Journal Publication published 21 Apr, 2025 Read the published version in Nature Communications → Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-4536158","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":319177752,"identity":"9ca351d0-2dad-42be-ab2a-7ba13eea7bb9","order_by":0,"name":"Jinze Liu","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAqUlEQVRIiWNgGAWjYBACPmYGBoPEPzYQHg8xWthAWh42pJGiBYgZHzYcJkULO49BQeKO84n9MxIYH7xtI8phPAYGiWduJ864kcBsOJdoLQlst40NJBLYpHlJ0HIOpIX9N/FaEtsOyIFsYSZSC1uBQcKZZDmJMw+bJeecI0ILP//hbYY/Kux4+NuTD354U0aEFpBFBhCasYE49UDA/IBopaNgFIyCUTAyAQCR7S03aCl0DwAAAABJRU5ErkJggg==","orcid":"","institution":"Virginia Commonwealth University","correspondingAuthor":true,"prefix":"","firstName":"Jinze","middleName":"","lastName":"Liu","suffix":""},{"id":319177753,"identity":"124d97de-ba37-4137-8ca4-c769365db349","order_by":1,"name":"Khoa Huynh","email":"","orcid":"","institution":"Virginia Commonwealth University","correspondingAuthor":false,"prefix":"","firstName":"Khoa","middleName":"","lastName":"Huynh","suffix":""},{"id":319177754,"identity":"1c6685a6-88b7-4e74-a326-2dc2878cf54d","order_by":2,"name":"Katarzyna Tyc","email":"","orcid":"https://orcid.org/0000-0002-3850-6934","institution":"Virginia Commonwealth University","correspondingAuthor":false,"prefix":"","firstName":"Katarzyna","middleName":"","lastName":"Tyc","suffix":""},{"id":319177755,"identity":"3f16b07a-bb92-4f72-8750-c0d08e527897","order_by":3,"name":"Bruno Fernandes Matuck","email":"","orcid":"","institution":"Lab of Oral \u0026 Craniofacial Innovation (LOCI)","correspondingAuthor":false,"prefix":"","firstName":"Bruno","middleName":"Fernandes","lastName":"Matuck","suffix":""},{"id":319177756,"identity":"6c0ed12b-748b-41e0-b6f0-afb7b863c51f","order_by":4,"name":"Quinn Easter","email":"","orcid":"https://orcid.org/0000-0002-8018-0405","institution":"Lab of Oral \u0026 Craniofacial Innovation (LOCI)","correspondingAuthor":false,"prefix":"","firstName":"Quinn","middleName":"","lastName":"Easter","suffix":""},{"id":319177757,"identity":"f1689f3f-4bb6-489a-8224-22a9ff5c8233","order_by":5,"name":"Aditya Pratapa","email":"","orcid":"","institution":"Duke University","correspondingAuthor":false,"prefix":"","firstName":"Aditya","middleName":"","lastName":"Pratapa","suffix":""},{"id":319177758,"identity":"57fff7b7-433f-4c8a-a3dc-275989c6665e","order_by":6,"name":"Nikhil Kumar","email":"","orcid":"","institution":"Lab of Oral \u0026 Craniofacial Innovation (LOCI)","correspondingAuthor":false,"prefix":"","firstName":"Nikhil","middleName":"","lastName":"Kumar","suffix":""},{"id":319177759,"identity":"0ba8991b-67db-463a-a1a5-1fa0ea447b69","order_by":7,"name":"Paola Pérez","email":"","orcid":"https://orcid.org/0000-0002-1695-5763","institution":"Salivary Disorders Unit, National Institute of Dental and Craniofacial Research, National Institutes of Health, Bethesda","correspondingAuthor":false,"prefix":"","firstName":"Paola","middleName":"","lastName":"Pérez","suffix":""},{"id":319177760,"identity":"b0d1fbaf-d2d3-4444-80e2-d5afbebe0c8b","order_by":8,"name":"Rachel Kulchar","email":"","orcid":"https://orcid.org/0000-0002-2904-7107","institution":"Salivary Disorders Unit, National Institute of Dental and Craniofacial Research, National Institutes of Health, Bethesda","correspondingAuthor":false,"prefix":"","firstName":"Rachel","middleName":"","lastName":"Kulchar","suffix":""},{"id":319177761,"identity":"70ab414e-8fe3-4075-afe3-4c970b33ed3c","order_by":9,"name":"Thomas Pranzatelli","email":"","orcid":"","institution":"Adeno-Associated Virus Biology Section, National Institute of Dental and Craniofacial Research, National Institutes of Health","correspondingAuthor":false,"prefix":"","firstName":"Thomas","middleName":"","lastName":"Pranzatelli","suffix":""},{"id":319177762,"identity":"801244a1-cc59-4353-95c4-15deac74581c","order_by":10,"name":"Deiziane Souza","email":"","orcid":"","institution":"Medicine School of University of Sao Paulo","correspondingAuthor":false,"prefix":"","firstName":"Deiziane","middleName":"","lastName":"Souza","suffix":""},{"id":319177763,"identity":"da8f0fa9-d751-4ea4-8402-31e5a4f78c31","order_by":11,"name":"Theresa Weaver","email":"","orcid":"","institution":"Lab of Oral \u0026 Craniofacial Innovation (LOCI)","correspondingAuthor":false,"prefix":"","firstName":"Theresa","middleName":"","lastName":"Weaver","suffix":""},{"id":319177764,"identity":"bce8674a-7f8b-4889-b1dd-49679a417968","order_by":12,"name":"Xufeng Qu","email":"","orcid":"","institution":"Massey Cancer Center","correspondingAuthor":false,"prefix":"","firstName":"Xufeng","middleName":"","lastName":"Qu","suffix":""},{"id":319177765,"identity":"919dd379-2041-4e46-840a-69de412cad6e","order_by":13,"name":"Luiz Soares Junior","email":"","orcid":"","institution":"Division of Dentistry of Hospital das Clinicas of University of Sao Paulo","correspondingAuthor":false,"prefix":"","firstName":"Luiz","middleName":"Soares","lastName":"Junior","suffix":""},{"id":319177766,"identity":"d2262388-eef6-4b84-b751-cd8db0e0b2ad","order_by":14,"name":"Marisa Dolhnokoff","email":"","orcid":"","institution":"Medicine School of University of Sao Paulo","correspondingAuthor":false,"prefix":"","firstName":"Marisa","middleName":"","lastName":"Dolhnokoff","suffix":""},{"id":319177767,"identity":"aa66476b-6fcd-497d-9670-dba65ef27f57","order_by":15,"name":"David Kleiner","email":"","orcid":"https://orcid.org/0000-0003-3442-4453","institution":"Center for Cancer Research, National Cancer Institute, National Institutes of Health","correspondingAuthor":false,"prefix":"","firstName":"David","middleName":"","lastName":"Kleiner","suffix":""},{"id":319177768,"identity":"39dedae7-81f1-416c-bac4-4864f89cb238","order_by":16,"name":"Stephen Hewitt","email":"","orcid":"https://orcid.org/0000-0001-8283-1788","institution":"National Institutes of Health","correspondingAuthor":false,"prefix":"","firstName":"Stephen","middleName":"","lastName":"Hewitt","suffix":""},{"id":319177769,"identity":"2d6dbcdb-7d36-43e5-82e3-f31cce3256b0","order_by":17,"name":"Luiz da Silva","email":"","orcid":"","institution":"Medicine School of University of Sao Paulo","correspondingAuthor":false,"prefix":"","firstName":"Luiz","middleName":"da","lastName":"Silva","suffix":""},{"id":319177770,"identity":"83c66206-6ff7-4765-97cd-cfa623be1dcc","order_by":18,"name":"Vanderson Rocha","email":"","orcid":"","institution":"University of Sao Paulo","correspondingAuthor":false,"prefix":"","firstName":"Vanderson","middleName":"","lastName":"Rocha","suffix":""},{"id":319177771,"identity":"6cf14d9c-ae05-4a01-b511-09d101903783","order_by":19,"name":"Blake Warner","email":"","orcid":"https://orcid.org/0000-0002-4961-018X","institution":"Salivary Disorders Unit, National Institute of Dental and Craniofacial Research, National Institutes of Health, Bethesda","correspondingAuthor":false,"prefix":"","firstName":"Blake","middleName":"","lastName":"Warner","suffix":""},{"id":319177772,"identity":"af98dfa4-b3c3-4222-ba50-84f201e0202f","order_by":20,"name":"Kevin Byrd","email":"","orcid":"https://orcid.org/0000-0002-5565-0524","institution":"Lab of Oral \u0026 Craniofacial Innovation (LOCI)","correspondingAuthor":false,"prefix":"","firstName":"Kevin","middleName":"","lastName":"Byrd","suffix":""}],"badges":[],"createdAt":"2024-06-05 20:50:22","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-4536158/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-4536158/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1038/s41467-025-58874-4","type":"published","date":"2025-04-21T04:00:00+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":59153087,"identity":"6b2035e2-66ed-4e0e-a1e6-e6b8eaaf005f","added_by":"auto","created_at":"2024-06-27 03:05:35","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":563619,"visible":true,"origin":"","legend":"\u003cp\u003eGeneral TACIT Workflow: (a) Multiplex imaging employs both spatial proteomics (top) and spatial transcriptomics (bottom). After segmentation (b top), a CELLxFEATURE matrix is generated (c). Hierarchical cell type structures (b bottom) are formulated based on panel design, expert knowledge, and scRNA-seq marker matching, resulting in a CELLTYPExMARKER matrix (c). Cells are organized into microclusters (MCs) by a community-based Louvain algorithm, averaging 0.1%-0.5% of the population (d top). These matrices are then used to compute Cell Type Relevance (CTR) scores for all cell types across cells (d bottom). Optimal thresholds are established to classify cells as clean if they meet one threshold or mixed if multiple (e). The UMAP with all features shows no clear separation between two distinct cell types (g – top left); however, clear segregation appears when only relevant features are used in the UMAP embedding (g – top right). Mixed identities are resolved by analyzing the mode of cell types within their k-nearest neighbors (g – bottom). Validation is performed via heatmaps comparing mean marker and cell type values with the CELLTYPExMARKER matrix (h – top), and by calculating enrichment scores for each cell type (i – bottom). The UMAP plot illustrates spatial distributions with cell type annotations (j top-right) and connections of cell type clusters (j bottom-left), combining cell type and state analyses (j bottom-right). Extended details of step e: Threshold derivation extends to segmental regression on ordered median CTR scores across all MCs to identify breakpoints (i \u0026amp; ii), defining “low relevance group (LRG)” and “high relevance group (HRG)” (ii). The determined CTR threshold minimizes classification error, distinguishing between LRG and HRG (iv \u0026amp; v). Cells above the threshold are highlighted in red on the UMAP, while those below are in grey (vi).\u003c/p\u003e","description":"","filename":"Figure1.png","url":"https://assets-eu.researchsquare.com/files/rs-4536158/v1/f41e0994bcf24776bda29b85.png"},{"id":59153090,"identity":"34562e37-d7cf-4532-9c11-5ac5b96eafc3","added_by":"auto","created_at":"2024-06-27 03:05:36","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":882954,"visible":true,"origin":"","legend":"\u003cp\u003eApplication of TACIT on PhenoCycler data from PCF-CRC (top panel) and PCF-HI (bottom panel). (a,g) Examples of spatial plots color-coded by identified cell types, illustrating the spatial distribution and clustering of cells as determined by TACIT. These plots demonstrate how TACIT preserves the spatial structure of cell types, maintaining consistency with the reference data. (e,k) UMAP representations with cell type delineations, showing the clustering of cells in a two-dimensional space. TACIT's UMAP plots reveal a higher degree of similarity to the reference clusters compared to other methods, indicating its superior performance in accurately identifying cell types. (f,i) Heatmaps comparing the mean marker values for each cell type identified by TACIT and other existing methods. TACIT's heatmaps exhibit distinct and clear unique marker expressions for each cell type, with a diagonal pattern that highlights its precise cell type identification capabilities. (d,j) Recall, precision, and F1 score comparisons between TACIT (PCF-CRC: 0.74 (Recall), 0.79 (Precision), 0.75 (F1), PCF-HI: 0.73 (Recall), 0.79 (Precision), 0.75 (F1)) and existing methods, benchmarked against the reference. TACIT consistently outperforms other methods, achieving higher recall, precision, and F1 scores, which underscores its accuracy and reliability in cell type identification. (e,k) Correlation plots illustrating the relationships between different cell type identification methods for both abundant cell types and rare cell types. TACIT shows strong correlations with the reference data, particularly for rare cell types (PCF-CRC: R=0.58, PCF-HI: R=0.76), where it demonstrates a higher degree of similarity in cell type identification compared to other methods. (f,l) Intensity comparison of unique markers between TACIT and existing methods. TACIT displays significantly different enrichment scores, particularly when compared to methods like Louvain (PCF-CRC \u0026amp; PCF-HI: p-value\u0026lt;0.05) or SCINA (PCF-CRC: p-value\u0026lt;0.05), indicating its enhanced ability to identify and distinguish unique cell markers.\u003c/p\u003e","description":"","filename":"Figure2.png","url":"https://assets-eu.researchsquare.com/files/rs-4536158/v1/df3da1b1ae71b0dbe9686663.png"},{"id":59153096,"identity":"6f2d03cd-8827-40d9-8d25-9e4c57e65bf2","added_by":"auto","created_at":"2024-06-27 03:05:37","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":718183,"visible":true,"origin":"","legend":"\u003cp\u003eApplication of TACIT on Xenium data. (a) UMAP and (b) spatial plots color-coded by identified cell types. The UMAP plots demonstrate TACIT's ability to cluster cells accurately, showing a clear separation of different cell types. Epithelial such as mucous acinar, myoepithelial, and seromucous acinar cells form more distinct and clear clusters under TACIT's annotation compared to Louvain and Seurat Transfer methods. The spatial plots further illustrate the spatial distribution of these cell types, maintaining the structural integrity and spatial organization consistent with the reference data. (c) Heatmaps depicting cell types and markers between TACIT, Louvain, Seurat transfer, and the signature matrix. TACIT's heatmaps present clear and distinct patterns, highlighting its precise identification of cell types and markers. This clarity is especially notable when compared to the other methods, which show less distinct marker expressions. (de) UMAP plots with low granularity cell types across the three methods. TACIT's enhanced capabilities are further exemplified by its identification of rare and diverse cell types, such as duct cells and duct progenitors, as well as various T cell types including CD4, CD8, CD8 exhausted, and T cell progenitors. (f) Correlation plot of cell type proportions between the three methods in Xenium, compared with scRNA cell type proportions. TACIT shows a higher correlation (Spearman Correlation, R=0.84) with scRNA cell type proportions, indicating a more consistent and reliable identification of cell types. In contrast, Seurat transfer and Louvain show lower correlations of 0.49 and 0.69, respectively. (g) TACIT and Seurat transfer able to find all the cell type matches with scRNA. (h-i) Intensity comparison of unique markers between TACIT and existing methods. TACIT exhibits a higher intensity of unique marker expressions compared to Louvain, with a log2 fold change (p-value\u0026lt;0.05), and shows significant performance over Louvain and Seurat transfer, with a -log10 adjusted p-value (p-value\u0026lt;0.05).\u003c/p\u003e","description":"","filename":"Figure3.png","url":"https://assets-eu.researchsquare.com/files/rs-4536158/v1/29d4d7fd656db854e78ad720.png"},{"id":59153088,"identity":"01456f12-0ea5-4215-865a-239a9cf3525d","added_by":"auto","created_at":"2024-06-27 03:05:36","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":1338341,"visible":true,"origin":"","legend":"\u003cp\u003eSingle-Slide Spatial Multiomics Annotation using TACIT (a) A spatial transcriptomics experiment on minor salivary glands from GVHD patients used a Xenium platform with a 280-gene panel targeting structural and immune cells, revealing a high-density immune area with overlays of specific transcripts. (b) A subsequent spatial proteomics experiment on the same slide utilized a Phenocycler Fusion with a 36-antibody panel, sharing the segmentation mask for consistent spatial single-cell data extraction. (c) UMAP analysis of the Xenium data with TACIT and Louvain showed greater annotation granularity with TACIT, highlighting cell types identified only by TACIT (arrows). (d) A Voronoi plot for a GVHD case displayed detailed annotation reconstruction by TACIT, showing the heterogeneity in a dense immune infiltrate. (e) A Venn diagram demonstrated that TACIT identified 22 cell types, including four not matched by Louvain, although Louvain’s detected types were also identified by TACIT. (f) The absolute error in cell type assignments compared to human pathologist evaluations varied between TACIT and Louvain. (g) Another UMAP from the Phenocycler Fusion data emphasized TACIT's higher granularity, with unique cell types marked (arrows). (h) A second Voronoi plot based on spatial proteomics data for a GVHD case illustrated TACIT’s annotation at a slightly lower resolution than the transcriptomics data. (i) A proteomics Venn diagram showed TACIT recognized and assigned 18 cell types, with two structural and two immune types uniquely detected. (j) The absolute error in cell quantity signatures from a spatial transcriptomics assay, compared with a human pathologist’s evaluation for each cell type, confirmed TACIT's precision over Louvain.\u003c/p\u003e","description":"","filename":"Figure4.png","url":"https://assets-eu.researchsquare.com/files/rs-4536158/v1/98c509c388ecfb9a4dff70bb.png"},{"id":59153099,"identity":"d3d46ecf-06ac-4775-a47c-69a5babd00b1","added_by":"auto","created_at":"2024-06-27 03:05:39","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":1254830,"visible":true,"origin":"","legend":"\u003cp\u003eApplication of TACIT in a Multimodal Single-Slide Tertiary Lymphoid Structure (a) Spatial transcriptomics and proteomics assays utilize segmentation to extract single-cell data, transferring the segmentation mask between experiments. However, this can lead to marker bleed-through; in proteomics, immunofluorescence markers stain the edges of adjacent B cells. In transcriptomics, probes such as the MS4A1 gene are found outside B cell boundaries in a GVHD minor salivary gland's TLS. (b) TACIT and Louvain exhibit varying performances when analyzing high-density immune areas like a TLS. TACIT identifies a more detailed and expected population of immune cells within the TLS compared to Louvain. (c) A heatmap displays the genes and proteins used to create cell signatures by TACIT and Louvain. Despite using the same list of genes, TACIT outperforms Louvain by providing clearer markers for each cell type and more precise cell recognition in high-density immune areas. (d) Voronoi plots demonstrate how different cell assignments lead to varied analysis outcomes. TACIT's reconstruction reveals a diverse mix of immune cells, small vessels, and antigen-presenting cells typical of a TLS. In contrast, Louvain shows lower resolution, merging all immune cells into broad categories of one innate and one adaptive type. (e) The choice of tools for cell assignment in multi-omics spatial assays impacts downstream analysis. The neighborhood analysis with TACIT illustrates expected cell proximities in a TLS, showing B cells and dendritic cells near small vessels and T cells. Conversely, Louvain shows unilateral interactions, focusing solely on the most abundant structural cell types in the analyzed ROI. (f) Using a single slide for spatial proteomics and transcriptomics allows for the identification of cell types and the assignment of specific biomolecules like chemokines, interleukins, and immune checkpoints to cells. This method not only reveals cellular patterns but also aids in studying spatial cell-cell communication. ROI reconstruction with TACIT assigned CD247 to T cells, B cells, and macrophages, highlighting diverse interactions. Conversely, the Clustering signature was exclusive to B cells, concentrated around capillaries, with no further interactions.\u003c/p\u003e","description":"","filename":"Figure5.png","url":"https://assets-eu.researchsquare.com/files/rs-4536158/v1/2ef33e7fe8a6c83dec1390e2.png"},{"id":59153089,"identity":"4163a063-74ee-473b-bc3d-8658da385562","added_by":"auto","created_at":"2024-06-27 03:05:36","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":933636,"visible":true,"origin":"","legend":"\u003cp\u003eMultimodal analysis using ST and SP in a single slide. (a) Two assays were combined on the same slide and section: Phenocycler Fusion (SP) and Xenium (ST). A segmentation mask was created using a human-in-the-loop approach and inputted into the Xenium Ranger. This mask was then transferred to the SP assay, maintaining cell IDs between the two experiments.(b) After segmentation, a matrix was extracted containing the pixel values of each immunofluorescent channel from the SP and the transcripts per cell from the ST. (c) This cell-by-feature matrix was then normalized and cell-assigned using TACIT. (d). The matched number of cells assigned by the SP and ST assays was quantified to evaluate the correlation in cell assignment for each major cell type – structural and immune cells. The correlation for structural cells using all transcripts and proteins was 0.37, and for immune cells, it was 0.01. (e). After the initial annotation, specific cell markers were used to assign cell types that had both protein and transcript designations in the proteomics and transcriptomics assays. The masks of cells annotated in three different ROIs with a high density of immune cells showed 34% agreement when using all markers. (f). A smaller subset of matched protein and RNA panels was utilized to improve agreement. The Voronoi mask showed better convergence in cell type annotation, increasing cell ID matching to 81%. (g-h) The difference in annotation by each approach for each of the six cell types selected using matched protein and RNA markers showed an improvement in cell assignment, with the proportion of the cell types. (i). After multimodal cell assignment, TACIT was also able to provide cell state markers for each cell. PD-1 and PDCD1 were used to understand the ratio of transcripts and proteins in high-density immune cell ROIs. The presence of these two markers was analyzed using SP alone, ST alone, and the two assays combined. (j) The proportion of positivity cell state in mRNA such as PDCD1 and MKI67 are significantly lower than PD-1 (p-value\u0026lt;0.05) and Ki67 (p-value\u0026lt;0.05) in protein for B cells and CD4+ T cells across TLS.\u003c/p\u003e","description":"","filename":"Figure6.png","url":"https://assets-eu.researchsquare.com/files/rs-4536158/v1/63de6ce0678587a033019fb0.png"},{"id":81090428,"identity":"4d9b25be-3c2e-4215-8ced-6204dbc1127f","added_by":"auto","created_at":"2025-04-22 07:05:57","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1516850,"visible":true,"origin":"","legend":"","description":"","filename":"TACITFINALNG.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4536158/v1_covered_4ee4557b-da24-495d-9c19-c29cdfe17522.pdf"},{"id":59153092,"identity":"4bbe2d3d-6010-456d-8531-70a822934e94","added_by":"auto","created_at":"2024-06-27 03:05:37","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":717012,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cbr\u003e\u003c/p\u003e","description":"","filename":"Extended1.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4536158/v1/eb0ad95198f0242579ed9b3d.pdf"},{"id":59154232,"identity":"2d76eefe-ad34-4b27-9d76-e678541b08f8","added_by":"auto","created_at":"2024-06-27 03:21:37","extension":"pdf","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":1105706,"visible":true,"origin":"","legend":"","description":"","filename":"Extended2.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4536158/v1/ba9ec61912a17a5a4b80435c.pdf"},{"id":59153101,"identity":"ed63dd33-2173-4e1f-a130-594780051262","added_by":"auto","created_at":"2024-06-27 03:05:39","extension":"pdf","order_by":3,"title":"","display":"","copyAsset":false,"role":"supplement","size":9748819,"visible":true,"origin":"","legend":"","description":"","filename":"Extended3.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4536158/v1/b011335c06a8fdbc82e643b0.pdf"},{"id":59153098,"identity":"f24231fe-d715-448d-a75b-b9794518413b","added_by":"auto","created_at":"2024-06-27 03:05:37","extension":"pdf","order_by":4,"title":"","display":"","copyAsset":false,"role":"supplement","size":1686240,"visible":true,"origin":"","legend":"","description":"","filename":"Extended4.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4536158/v1/df2d145498718b54059bbf4b.pdf"},{"id":59153097,"identity":"64467a31-33dd-4a17-9e5c-024c61397a8a","added_by":"auto","created_at":"2024-06-27 03:05:37","extension":"pdf","order_by":5,"title":"","display":"","copyAsset":false,"role":"supplement","size":3285657,"visible":true,"origin":"","legend":"","description":"","filename":"Extended5.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4536158/v1/b473de8959a4c19b603df41a.pdf"},{"id":59153100,"identity":"20f18f62-ed93-4398-a646-a4d681a36d26","added_by":"auto","created_at":"2024-06-27 03:05:39","extension":"pdf","order_by":6,"title":"","display":"","copyAsset":false,"role":"supplement","size":28383128,"visible":true,"origin":"","legend":"","description":"","filename":"Extended6.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4536158/v1/8bb3884c2eec772cc37f62ed.pdf"},{"id":59153738,"identity":"fbd10e5e-8c56-4981-beec-62dc5430cd57","added_by":"auto","created_at":"2024-06-27 03:13:37","extension":"pdf","order_by":7,"title":"","display":"","copyAsset":false,"role":"supplement","size":7868406,"visible":true,"origin":"","legend":"","description":"","filename":"Extended7.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4536158/v1/8018a240f94d614868e19dee.pdf"},{"id":59153094,"identity":"85d1b0ae-100a-484d-9688-7eee4b53d2f8","added_by":"auto","created_at":"2024-06-27 03:05:37","extension":"xlsx","order_by":8,"title":"","display":"","copyAsset":false,"role":"supplement","size":28844,"visible":true,"origin":"","legend":"","description":"","filename":"ExtendedData8.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-4536158/v1/003f4875e4470bac2e3f648d.xlsx"},{"id":59153091,"identity":"3b7b7745-e9d9-4015-b46e-d0751709e86e","added_by":"auto","created_at":"2024-06-27 03:05:36","extension":"pdf","order_by":9,"title":"","display":"","copyAsset":false,"role":"supplement","size":407404,"visible":true,"origin":"","legend":"\u003cp\u003eEditorial Policy Checklist\u003c/p\u003e","description":"","filename":"nreditorialpolicychecklistflatTACIT.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4536158/v1/741c7cd3962ac4acf9e10cc8.pdf"}],"financialInterests":"\u003cb\u003eYes\u003c/b\u003e there is potential Competing Interest.\nThe authors had access to the study data and reviewed and approved the final manuscript. Although the authors view each of these as noncompeting financial interests, KMB, QTE, BFM, and BMW are all active members of the Human Cell Atlas; furthermore, KMB is a scientific advisor at Arcato Laboratories and Orange Grove Bio. All other authors declare no competing interests.","formattedTitle":"Spatial Deconvolution of Cell Types and Cell States at Scale Utilizing TACIT","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":true,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":true,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"nature-portfolio","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"","title":"Nature Portfolio","twitterHandle":"","acdcEnabled":false,"dfaEnabled":false,"editorialSystem":"ejp","reportingPortfolio":"","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"spatial biology, multimodal, transcriptomics, proteomics, artificial intelligence, machine learning, deep learning, multiplex imaging, fluorescence microscopy, brain, intestine, salivary gland, cell typing, single cell analysis, spatial multiomics","lastPublishedDoi":"10.21203/rs.3.rs-4536158/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-4536158/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"Identifying cell types and states remains a time-consuming, error-prone challenge for spatial biology. While deep learning is increasingly used, it is difficult to generalize due to variability at the level of cells, neighborhoods, and niches in health and disease. To address this, we developed TACIT, an unsupervised algorithm for cell annotation using predefined signatures that operates without training data. TACIT uses unbiased thresholding to distinguish positive cells from background, focusing on relevant markers to identify ambiguous cells in multiomic assays. Using five datasets (5,000,000-cells; 51-cell types) from three niches (brain, intestine, gland), TACIT outperformed existing unsupervised methods in accuracy and scalability. Integrating TACIT-identified cell types with a novel Shiny app revealed new phenotypes in two inflammatory gland diseases. Finally, using combined spatial transcriptomics and proteomics, we discovered under- and overrepresented immune cell types and states in regions of interest, suggesting multimodality is essential for translating spatial biology to clinical applications.","manuscriptTitle":"Spatial Deconvolution of Cell Types and Cell States at Scale Utilizing TACIT","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-06-27 03:05:30","doi":"10.21203/rs.3.rs-4536158/v1","editorialEvents":[],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"nature-communications","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"NCOMMS","sideBox":"Learn more about [Nature Communications](http://www.nature.com/ncomms/)","snPcode":"","submissionUrl":"https://mts-ncomms.nature.com/","title":"Nature Communications","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"ejp","reportingPortfolio":"Nature Communications","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"bf048cc9-eed2-47bc-9287-4cdb5a1594ef","owner":[],"postedDate":"June 27th, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[{"id":33747764,"name":"Biological sciences/Computational biology and bioinformatics/Computational models"},{"id":33747765,"name":"Biological sciences/Computational biology and bioinformatics/Statistical methods"}],"tags":[],"updatedAt":"2025-04-22T07:05:47+00:00","versionOfRecord":{"articleIdentity":"rs-4536158","link":"https://doi.org/10.1038/s41467-025-58874-4","journal":{"identity":"nature-communications","isVorOnly":false,"title":"Nature Communications"},"publishedOn":"2025-04-21 04:00:00","publishedOnDateReadable":"April 21st, 2025"},"versionCreatedAt":"2024-06-27 03:05:30","video":"","vorDoi":"10.1038/s41467-025-58874-4","vorDoiUrl":"https://doi.org/10.1038/s41467-025-58874-4","workflowStages":[]},"version":"v1","identity":"rs-4536158","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-4536158","identity":"rs-4536158","version":["v1"]},"buildId":"qtupq5eGEP_6zYnWcrvyt","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}