pith. sign in

arxiv: 2602.10644 · v1 · pith:CEM6A23Fnew · submitted 2026-02-11 · 🧬 q-bio.BM

Towards Universal Spatial Transcriptomics Super-Resolution: A Generalist Physically Consistent Flow Matching Framework

Xinlei Huang , Weihao Dai , Zijun Qin , Xin Yu , Di Wang , Yanran Liu , Lixin Cheng , Xubin Zheng This is my paper

Pith reviewed 2026-05-21 13:56 UTC · model grok-4.3

classification 🧬 q-bio.BM
keywords spatial transcriptomicssuper-resolutionflow matchingphysical consistencygeneralist modelzero-shot generalizationgene expressiontissue heterogeneity
0
0 comments X

The pith

SRast super-resolves spatial transcriptomics data by decoupling gene semantics from spatial geometry and enforcing mass conservation via flow matching.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper presents SRast as a generalist framework that takes low-resolution spatial transcriptomics inputs and produces higher-resolution outputs while handling biological differences across samples. It separates gene expression patterns from the spatial layout of the tissue, aligns these parts through self-supervised training, and then models the resolution boost as a flow that preserves total gene counts in local regions. This setup is tested across many species, tissues, and measurement platforms to show it works without retraining on new data and produces structures that respect physical rules. A reader would care because current high-resolution methods are costly and limited, so a reliable way to enhance cheaper low-resolution data could expand access to detailed tissue maps.

Core claim

SRast reformulates spatial transcriptomics super-resolution as ratio prediction on the simplex and solves it with a flow matching model that learns optimal transport transformations between low- and high-resolution distributions; the model is built on a decoupling architecture that separates gene semantics representation from spatial geometry deconvolution and aligns the resulting latent spaces with self-supervised learning to reduce shifts caused by biological heterogeneity.

What carries the argument

SRast's decoupling architecture combined with a flow matching model for ratio prediction on the simplex, which learns geometric transformations that enforce local mass conservation.

If this is right

  • The method produces super-resolved maps that conserve local gene mass, allowing direct comparison of total expression levels before and after enhancement.
  • Zero-shot application to new tissues becomes feasible because the latent alignment reduces the impact of sample-specific biological variation.
  • Fine-grained structures such as cell-type boundaries and spatial gradients can be recovered without platform-specific retraining.
  • The framework applies uniformly across species and measurement technologies, reducing the need for custom models per experiment.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Similar decoupling and flow-based constraints could be tested on other spatial omics modalities such as proteomics or metabolomics to check transferability.
  • The mass-conservation property might enable direct fusion of super-resolved transcriptomics with lower-resolution but higher-coverage bulk sequencing data.
  • If the alignment step proves stable, the same architecture could support temporal super-resolution when tracking tissue changes over time.

Load-bearing premise

Self-supervised learning on decoupled gene semantics and spatial geometry representations can reliably align latent distributions and mitigate cross-sample shifts caused by inherent biological heterogeneity.

What would settle it

Running SRast on a held-out dataset from a new species or platform and finding that the output gene counts in local neighborhoods deviate from the input totals or that performance drops below existing methods would show the physical consistency and generalization claims do not hold.

Figures

Figures reproduced from arXiv: 2602.10644 by Di Wang, Lixin Cheng, Weihao Dai, Xinlei Huang, Xin Yu, Xubin Zheng, Yanran Liu, Zijun Qin.

Figure 1
Figure 1. Figure 1: Overview of our proposed SRast framework. (Left) Structure-Aware Semantic Alignment: The model employs a [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 3
Figure 3. Figure 3: Comparison of Running Time Cost between SRast [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figure 2
Figure 2. Figure 2: Umap visualization comparison of PCA features (a), [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗
read the original abstract

Spatial transcriptomics provides an unprecedented perspective for deciphering tissue spatial heterogeneity. However, high-resolution spatial transcriptomic technology remains constrained by limited gene coverage, technical complexity, and high cost. Existing spatial transcriptomics super-resolution methods from low resolution data suffer from two fundamental limitations: poor out-of-distribution generalization stemming from a neglect of inherent biological heterogeneity, and a lack of physical consistency. To address these challenges, we propose SRast, a novel physically constrained generalist framework designed for robust spatial transcriptomics super-resolution. To tackle heterogeneity, SRast employs a strategic decoupling architecture that explicitly decouples gene semantics representation from spatial geometry deconvolution, utilizing self-supervised learning to align latent distributions and mitigate cross-sample shifts. Regarding physical priors, SRast reformulates the task as ratio prediction on the simplex, performing a flow matching model to learn optimal transport-based geometric transformations that strictly enforce local mass conservation. Extensive experiments across diverse species, tissues, and platforms demonstrate that SRast achieves state-of-the-art performance, exhibiting superior zero-shot generalization capabilities and ensuring physical consistency in recovering fine-grained biological structures.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The manuscript introduces SRast, a generalist framework for spatial transcriptomics super-resolution. It decouples gene semantics representations from spatial geometry deconvolution via self-supervised learning to mitigate cross-sample biological heterogeneity, and reformulates super-resolution as simplex ratio prediction solved by a flow-matching model that learns optimal-transport transformations asserted to enforce local mass conservation. The authors report state-of-the-art performance with superior zero-shot generalization across species, tissues, and platforms while recovering fine-grained structures in a physically consistent manner.

Significance. If the performance and conservation claims are substantiated with quantitative evidence, the work would offer a practically useful advance for the field: a single model that generalizes across heterogeneous spatial transcriptomics datasets without retraining while respecting mass-conservation priors that are biologically meaningful for gene-expression recovery. The explicit decoupling strategy and flow-matching formulation on the simplex are conceptually attractive and could reduce reliance on platform-specific high-resolution training data.

major comments (2)
  1. [Abstract / Results] Abstract and Results sections: the central SOTA and zero-shot generalization claims are stated without any reported quantitative metrics, baselines, error bars, or cross-validation details in the provided text. This absence prevents evaluation of whether the performance improvements are statistically meaningful or merely incremental.
  2. [Methods] Methods (flow-matching formulation): the claim that the model 'strictly enforces local mass conservation' via simplex ratio prediction and optimal-transport transformations is load-bearing for the 'physically consistent' part of the title and abstract. No auxiliary conservation loss, post-hoc projection step, or reported per-region mass-error statistics (e.g., total variation or simplex violation norms) are described; continuous-time flow matching plus discretization can introduce drift, so explicit verification is required.
minor comments (2)
  1. [Methods] Notation for the simplex ratio prediction and the precise definition of the flow-matching objective (continuous vs. discrete) should be clarified with an equation or pseudocode block to allow reproducibility.
  2. [Figures / Experiments] Figure captions and experimental tables should explicitly list the number of samples per species/tissue/platform and the exact train/test splits used for the zero-shot experiments.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive and insightful comments, which have helped us strengthen the presentation of our results and the substantiation of our claims. We address each major comment point by point below and have revised the manuscript accordingly.

read point-by-point responses

  1. Referee: [Abstract / Results] Abstract and Results sections: the central SOTA and zero-shot generalization claims are stated without any reported quantitative metrics, baselines, error bars, or cross-validation details in the provided text. This absence prevents evaluation of whether the performance improvements are statistically meaningful or merely incremental.

    Authors: The abstract is a concise summary and therefore omits specific numerical values, but the full Results section contains extensive quantitative evaluations, including direct comparisons against baselines, performance metrics with error bars from repeated runs, and details of the cross-validation protocol across multiple datasets. To improve immediate verifiability of the SOTA and zero-shot claims, we have revised the abstract to include representative quantitative highlights (e.g., average improvements in correlation and RMSE) while remaining within length limits, and we have added explicit pointers to the corresponding tables and figures. revision: yes

  2. Referee: [Methods] Methods (flow-matching formulation): the claim that the model 'strictly enforces local mass conservation' via simplex ratio prediction and optimal-transport transformations is load-bearing for the 'physically consistent' part of the title and abstract. No auxiliary conservation loss, post-hoc projection step, or reported per-region mass-error statistics (e.g., total variation or simplex violation norms) are described; continuous-time flow matching plus discretization can introduce drift, so explicit verification is required.

    Authors: We agree that explicit verification is valuable. The simplex-ratio formulation combined with flow matching on probability measures guarantees mass conservation by construction: every generated field stays within the probability simplex and the learned optimal-transport paths preserve total mass. To address discretization concerns, we have added (i) a brief theoretical note in Methods explaining why the continuous formulation plus the simplex constraint prevents drift, and (ii) quantitative per-region mass-error statistics (mean total-variation distance and maximum simplex-violation norm) in the revised Results, confirming errors remain below 10^{-3} across all evaluated tissues and platforms. revision: yes

Circularity Check

0 steps flagged

No circularity: claims rest on independent modeling choices rather than self-referential definitions or fitted outputs

full rationale

The paper's derivation chain is self-contained. It reformulates super-resolution as simplex ratio prediction and applies flow matching to learn optimal-transport transformations, asserting that this enforces local mass conservation by construction of the method. This is a standard modeling decision using established techniques (flow matching, self-supervised decoupling), not a reduction where the physical consistency result is defined in terms of the fitted outputs or inputs. No equations are shown that equate a prediction directly to a fitted parameter, no load-bearing self-citations appear in the abstract or claims, and the zero-shot generalization and SOTA performance are presented as empirical outcomes rather than tautological consequences of the inputs. The framework is externally falsifiable via conservation-error metrics on held-out data, satisfying the criteria for non-circular independent content.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The central claims rest on domain assumptions about biological heterogeneity being addressable via latent alignment and on the validity of local mass conservation as a physical prior for gene expression distributions; no free parameters or new entities are explicitly introduced in the abstract.

axioms (2)
  • domain assumption Biological heterogeneity across samples can be mitigated by self-supervised alignment of latent distributions after decoupling gene semantics from spatial geometry
    Invoked to justify the decoupling architecture for zero-shot generalization.
  • domain assumption Local mass conservation applies to gene expression distributions and can be strictly enforced by reformulating super-resolution as ratio prediction on the simplex via flow matching
    Directly supports the physical consistency guarantee.

pith-pipeline@v0.9.0 · 5741 in / 1313 out tokens · 51233 ms · 2026-05-21T13:56:34.538605+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

47 extracted references · 47 canonical work pages · 6 internal anchors

  1. [1]

    Patrik L Ståhl, Fredrik Salmén, Sanja Vickovic, Anna Lundmark, José Fernández Navarro, Jens Magnusson, Stefania Giacomello, Michaela Asp, Jakub O Westholm, Mikael Huss, et al. 2016. Visualization and analysis of gene expression in tissue sections by spatial transcriptomics.Science353, 6294 (2016), 78–82

    work page 2016

  2. [2]

    Chee-Huat Linus Eng, Michael Lawson, Qian Zhu, Ruben Dries, Noushin Koulena, Yodai Takei, Jina Yun, Christopher Cronin, Christoph Karp, Guo-Cheng Yuan, et al

    work page

  3. [3]

    Nature568, 7751 (2019), 235–239

    Transcriptome-scale super-resolved imaging in tissues by RNA seqFISH+. Nature568, 7751 (2019), 235–239

    work page 2019

  4. [4]

    Vivien Marx. 2021. Method of the Year: spatially resolved transcriptomics.Nature methods18, 1 (2021), 9–14

    work page 2021

  5. [5]

    Lambda Moses and Lior Pachter. 2022. Museum of spatial transcriptomics.Nature methods19, 5 (2022), 534–546

    work page 2022

  6. [6]

    Guangyao Mao, Yi Yang, Zhuojuan Luo, Chengqi Lin, and Peng Xie. 2024. Spa- tialQC: automated quality control for spatial transcriptome data.Bioinformatics 40, 8 (2024), btae458

    work page 2024

  7. [7]

    Edward Zhao, Matthew R Stone, Xing Ren, Jamie Guenthoer, Kimberly S Smythe, Thomas Pulliam, Stephen R Williams, Cedric R Uytingco, Sarah EB Taylor, Paul Nghiem, et al. 2021. Spatial transcriptomics at subspot resolution with BayesSpace. Nature biotechnology39, 11 (2021), 1375–1384

    work page 2021

  8. [8]

    Haiyan Wang, Yanping Zhang, Yangyang Zhang, Xuening Zhao, Zijia Bai, Xuejing Ma, and Chunguang Zhao. 2025. SpaVGN: A hybrid deep learning framework for high-resolution spatial transcriptomics data reconstruction and spatial domain identification.PLoS One20, 8 (2025), e0329122

    work page 2025

  9. [9]

    TN Kipf. 2016. Semi-supervised classification with graph convolutional networks. arXiv preprint arXiv:1609.02907(2016)

    work page internal anchor Pith review Pith/arXiv arXiv 2016

  10. [10]

    Daiwei Zhang, Amelia Schroeder, Hanying Yan, Haochen Yang, Jian Hu, Michelle YY Lee, Kyung S Cho, Katalin Susztak, George X Xu, Michael D Feldman, et al. 2024. Inferring super-resolution tissue architecture by integrating spatial transcriptomics with histology.Nature biotechnology42, 9 (2024), 1372–1377

    work page 2024

  11. [11]

    Amelia Schroeder, Melanie L Loth, Chunyu Luo, Sicong Yao, Hanying Yan, Daiwei Zhang, Sarbottam Piya, Edward Plowey, Wenxing Hu, Jean R Clemenceau, et al

    work page

  12. [12]

    Nature methods22, 9 (2025), 1911–1922

    Scaling up spatial transcriptomics for large-sized tissues: uncovering cellular-level tissue architecture beyond conventional platforms with iSCALE. Nature methods22, 9 (2025), 1911–1922

    work page 2025

  13. [13]

    Xiuyuan Wang, Fei Ye, and Xiao Liu. 2025. STRESS: Spatial Transcriptome Resolution Enhancing Method based on the State Space Model.bioRxiv(2025), 2025–06

    work page 2025

  14. [14]

    Tim Stuart and Rahul Satija. 2019. Integrative single-cell analysis.Nature reviews genetics20, 5 (2019), 257–272

    work page 2019

  15. [15]

    Hoa Thi Nhu Tran, Kok Siong Ang, Marion Chevrier, Xiaomeng Zhang, Nicole Yee Shin Lee, Michelle Goh, and Jinmiao Chen. 2020. A benchmark of batch-effect correction methods for single-cell RNA sequencing data.Genome biology21, 1 (2020), 12

    work page 2020

  16. [16]

    Sophia K Longo, Margaret G Guo, Andrew L Ji, and Paul A Khavari. 2021. In- tegrating single-cell and spatial transcriptomics to elucidate intercellular tissue dynamics.Nature Reviews Genetics22, 10 (2021), 627–644

    work page 2021

  17. [17]

    Dan Hendrycks and Kevin Gimpel. 2016. A baseline for detecting misclassified and out-of-distribution examples in neural networks.arXiv preprint arXiv:1610.02136 (2016)

    work page internal anchor Pith review Pith/arXiv arXiv 2016

  18. [18]

    Pang Wei Koh, Shiori Sagawa, Henrik Marklund, Sang Michael Xie, Marvin Zhang, Akshay Balsubramani, Weihua Hu, Michihiro Yasunaga, Richard Lanas Phillips, Irena Gao, et al. 2021. Wilds: A benchmark of in-the-wild distribution shifts. InInternational conference on machine learning. PMLR, 5637–5664

    work page 2021

  19. [19]

    2009.Digital image processing

    Rafael C Gonzalez. 2009.Digital image processing. Pearson education india

    work page 2009

  20. [20]

    Robert Keys. 2003. Cubic convolution interpolation for digital image processing. IEEE transactions on acoustics, speech, and signal processing29, 6 (2003), 1153– 1160

    work page 2003

  21. [21]

    1994.Kernel smoothing

    Matt P Wand and M Chris Jones. 1994.Kernel smoothing. CRC press

    work page 1994

  22. [22]

    Petar Velickovic, Guillem Cucurull, Arantxa Casanova, Adriana Romero, Pietro Lio, Yoshua Bengio, et al. 2017. Graph attention networks.stat1050, 20 (2017), 10–48550

    work page 2017

  23. [23]

    Diederik P Kingma and Max Welling. 2013. Auto-encoding variational bayes. arXiv preprint arXiv:1312.6114(2013)

    work page internal anchor Pith review Pith/arXiv arXiv 2013

  24. [24]

    Sergey Ioffe and Christian Szegedy. 2015. Batch normalization: Accelerating deep network training by reducing internal covariate shift. InInternational conference on machine learning. pmlr, 448–456

    work page 2015

  25. [25]

    William Peebles and Saining Xie. 2023. Scalable diffusion models with transform- ers. InProceedings of the IEEE/CVF international conference on computer vision. 4195–4205

    work page 2023

  26. [26]

    Geoffrey Hinton, Oriol Vinyals, and Jeff Dean. 2015. Distilling the knowledge in a neural network.arXiv preprint arXiv:1503.02531(2015)

    work page internal anchor Pith review Pith/arXiv arXiv 2015

  27. [27]

    Xingchao Liu, Chengyue Gong, and Qiang Liu. 2022. Flow straight and fast: Learning to generate and transfer data with rectified flow.arXiv preprint arXiv:2209.03003(2022)

    work page internal anchor Pith review Pith/arXiv arXiv 2022

  28. [28]

    Yaron Lipman, Ricky TQ Chen, Heli Ben-Hamu, Maximilian Nickel, and Matt Le

    work page

  29. [29]

    Flow matching for generative modeling.arXiv preprint arXiv:2210.02747 (2022)

    work page internal anchor Pith review Pith/arXiv arXiv 2022

  30. [30]

    Yahui Long, Kok Siong Ang, Raman Sethi, Sha Liao, Yang Heng, Lynn van Olst, Shuchen Ye, Chengwei Zhong, Hang Xu, Di Zhang, et al . 2024. Deciphering spatial domains from spatial multi-omics with SpatialGlue.Nature Methods21, 9 (2024), 1658–1667

    work page 2024

  31. [31]

    Sha Liao, Yang Heng, Weiqing Liu, Jinqiong Xiang, Yong Ma, Ligen Chen, Xiuwen Feng, Dongmei Jia, Diyan Liang, Caili Huang, et al . 2023. Integrated spatial transcriptomic and proteomic analysis of fresh frozen tissue based on stereo-seq. bioRxiv(2023), 2023–04

    work page 2023

  32. [32]

    Di Zhang, Yanxiang Deng, Petra Kukanja, Eneritz Agirre, Marek Bartosovic, Mingze Dong, Cong Ma, Sai Ma, Graham Su, Shuozhen Bao, et al. 2023. Spatial epigenome–transcriptome co-profiling of mammalian tissues.Nature616, 7955 (2023), 113–122

    work page 2023

  33. [33]

    Nir Ben-Chetrit, Xiang Niu, Ariel D Swett, Jesus Sotelo, Maria S Jiao, Caitlin M Stewart, Catherine Potenski, Paulius Mielinis, Patrick Roelli, Marlon Stoeckius, et al. 2023. Integration of whole transcriptome spatial profiling with protein markers.Nature biotechnology41, 6 (2023), 788–793

    work page 2023

  34. [34]

    Xuhua Yan, Kok Siong Ang, Lynn van Olst, Alex Edwards, Thomas Watson, Ruiqing Zheng, Rong Fan, Min Li, David Gate, and Jinmiao Chen. 2024. Mosaic integration of spatial multi-omics with SpaMosaic.BioRxiv(2024), 2024–10

    work page 2024

  35. [35]

    Fuqing Jiang, Xin Zhou, Yingying Qian, Miao Zhu, Li Wang, Zhuxia Li, Qingmei Shen, Minhan Wang, Fangfang Qu, Guizhong Cui, et al . 2023. Simultaneous profiling of spatial gene expression and chromatin accessibility during mouse brain development.Nature Methods20, 7 (2023), 1048–1057

    work page 2023

  36. [36]

    Mengnan Cheng, Liang Wu, Lei Han, Xin Huang, Yiwei Lai, Jiangshan Xu, Shuai Wang, Mei Li, Huiwen Zheng, Weimin Feng, et al. 2022. A cellular resolution spatial transcriptomic landscape of the medial structures in postnatal mouse brain.Frontiers in Cell and Developmental Biology10 (2022), 878346

    work page 2022

  37. [37]

    10x Genomics. 2022. Human Breast Cancer: Visium Fresh Frozen Whole Tran- scriptome. https://www.10xgenomics.com/resources/datasets/human-breast- cancer-visium-fresh-frozen-whole-transcriptome-1-standard

    work page 2022

  38. [38]

    10x Genomics. 2023. Visium CytAssist Gene and Protein Expres- sion Library of Human Glioblastoma, IF, 11mm (FFPE). https: //www.10xgenomics.com/datasets/gene-and-protein-expression-library- of-human-glioblastoma-cytassist-ffpe-2-standard

    work page 2023

  39. [39]

    10x Genomics. 2023. Gene and Protein Expression Library of Human Tonsil: CytAssist FFPE. https://www.10xgenomics.com/resources/datasets/gene-protein- expression-library-of-human-tonsil-cytassist-ffpe-2-standard

    work page 2023

  40. [40]

    10x Genomics. 2023. Visium CytAssist Gene and Protein Expression Library of Human Tonsil, H&E, 6.5 mm (FFPE). https://www.10xgenomics.com/datasets/ gene-protein-expression-library-of-human-tonsil-cytassist-ffpe-2-standard

    work page 2023

  41. [41]

    Maynard, Leonardo Collado-Torres, Lukas M

    Kristen R. Maynard, Leonardo Collado-Torres, Lukas M. Weber, Cedric Uytingco, Brianna K. Barry, Stephen R. Williams, Joseph L. Catallini II, Matthew N. Tran, Zachary Besich, Madhavi Tippani, Jennifer Chew, Yifeng Yin, Joel E. Kleinman, Thomas M. Hyde, Nikhil Rao, Stephanie C. Hicks, Keri Martinowich, and An- drew E. Jaffe. 2021. Transcriptome-scale spatia...

    work page doi:10.1038/s41593- 2021

  42. [42]

    Huuki-Myers, Abby Spangler, Nicholas J

    Louise A. Huuki-Myers, Abby Spangler, Nicholas J. Eagles, Kelsey D. Mont- gomergy, Sang Ho Kwon, Boyi Guo, Melissa Grant-Peters, Heena R. Divecha, Mad- havi Tippani, Chaichontat Sriworarat, Annie B. Nguyen, Prashanthi Ravichan- dran, Matthew N. Tran, Arta Seyedian, PsychENCODE Consortium, Thomas M. Hyde, Joel E. Kleinman, Alexis Battle, Stephanie C. Page,...

    work page doi:10.1126/science.adh1938 2024

  43. [43]

    Yahui Long, Kok Siong Ang, Mengwei Li, Kian Long Kelvin Chong, Raman Sethi, Chengwei Zhong, Hang XU, Zhiwei Ong, Karishma Sachaphibulkij, Ao Chen, et al. 2022. DeepST: A versatile graph contrastive learning framework for spatially informed clustering, integration, and deconvolution of spatial transcriptomics. (2022)

    work page 2022

  44. [44]

    Xiaoxiao Wang, David Venet, Frédéric Lifrange, Denis Larsimont, Mattia Rediti, Linnea Stenbeck, Floriane Dupont, Ghizlane Rouas, Andrea Joaquin Garcia, Ligia Craciun, et al. 2024. Spatial transcriptomics reveals substantial heterogeneity in triple-negative breast cancer with potential clinical implications.Nature communications15, 1 (2024), 10232

    work page 2024

  45. [45]

    J MacQueen. 1967. Multivariate observations. InProceedings ofthe 5th Berkeley Symposium on Mathematical Statisticsand Probability, Vol. 1. 281–297

    work page 1967

  46. [46]

    Vincent A Traag, Ludo Waltman, and Nees Jan Van Eck. 2019. From Louvain to Leiden: guaranteeing well-connected communities.Scientific reports9, 1 (2019), 1–12

    work page 2019

  47. [47]

    Luca Scrucca, Michael Fop, T Brendan Murphy, and Adrian E Raftery. 2016. mclust 5: clustering, classification and density estimation using Gaussian finite mixture models.The R journal8, 1 (2016), 289

    work page 2016