Skip to main content

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

Integrative spatial analysis of cell morphologies and transcriptional states with MUSE

Nature Biotechnology volume 40, pages 1200–1209 (2022)Cite this article

Subjects

A Publisher Correction to this article was published on 03 May 2022

This article has been updated

Abstract

Spatial transcriptomics enables the simultaneous measurement of morphological features and transcriptional profiles of the same cells or regions in tissues. Here we present multi-modal structured embedding (MUSE), an approach to characterize cells and tissue regions by integrating morphological and spatially resolved transcriptional data. We demonstrate that MUSE can discover tissue subpopulations missed by either modality as well as compensate for modality-specific noise. We apply MUSE to diverse datasets containing spatial transcriptomics (seqFISH+, STARmap or Visium) and imaging (hematoxylin and eosin or fluorescence microscopy) modalities. MUSE identified biologically meaningful tissue subpopulations and stereotyped spatial patterning in healthy brain cortex and intestinal tissues. In diseased tissues, MUSE revealed gene biomarkers for proximity to tumor region and heterogeneity of amyloid precursor protein processing across Alzheimer brain regions. MUSE enables the integration of multi-modal data to provide insights into the states, functions and organization of cells in complex biological tissues.

This is a preview of subscription content, access via your institution

Relevant articles

Open Access articles citing this article.

Access options

Rent or buy this article

Get just this article for as long as you need it

$39.95

Learn more

Prices may be subject to local taxes which are calculated during checkout

Fig. 1: Overview of MUSE and performance evaluation on simulated data.
Fig. 2: Evaluation of MUSE on seqFISH+ mouse cortex data.
Fig. 3: Analysis of MUSE clusters on STARmap cortex dataset.
Fig. 4: Application of MUSE to an ST PDAC dataset.
Fig. 5: Application of MUSE to a 10x Visium human intestine dataset.
Fig. 6: Application of MUSE to a multimodal AD dataset.

Data availability

seqFISH+ mouse cortex dataset: Transcript data were downloaded from the GitHub page of the seqFISH+ project (https://github.com/CaiGroup/seqFISH-PLUS) on 1 August 2019. Nissl and DAPI stained images were provided by the authors of the seqFISH+ paper.

STARmap mouse cortex dataset: Raw data were downloaded from the project page (http://clarityresourcecenter.org/) on 2 July 2019. Transcript profiles and cell segmentation masks were extracted from data using the Python pipeline provided by the authors at https://github.com/weallen/STARmap.

PDAC dataset: Both spatial transcriptomics (including gene expressions and H&E images) and scRNA-seq datasets were downloaded from the Gene Expression Omnibus (GEO) database with accession number GSE111672.

Intestine dataset: 10x Visium spatial transcriptomics were downloaded from the GEO database with accession number GSE158328.

AD dataset: Raw and normalized count matrix of the spatial transcriptomics were downloaded from the GEO database of the project (accession number GSE152506). Immunofluorescence images (Abeta, GFAP, NeuN and DAPI staining) that correspond to spatial transcriptomics data were downloaded from the ‘synapse.org’ page of the project (https://www.synapse.org/#!Synapse:syn22153884/wiki/603937) on 31 October 2020.

Code availability

Simulated tool for multi-modality data generation: Simulation code is available from GitHub (https://github.com/AltschulerWu-Lab/MUSE).

MUSE: MUSE is provided as a Python package under MIT license and can be installed through ‘pip install muse_sc’. Source code and demonstration code are available on GitHub (https://github.com/AltschulerWu-Lab/MUSE).

Change history

References

  1. Perlman, Z. E. et al. Multidimensional drug profiling by automated microscopy. Science 306, 1194–1198 (2004).

    Article  CAS  PubMed  Google Scholar 

  2. Loo, L.-H., Wu, L. F. & Altschuler, S. J. Image-based multivariate profiling of drug responses from single cells. Nat. Methods 4, 445–453 (2007).

    Article  CAS  PubMed  Google Scholar 

  3. Feldman, D. et al. Optical pooled screens in human cells. Cell 179, 787–799 (2019).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  4. Klein, A. M. et al. Droplet barcoding for single-cell transcriptomics applied to embryonic stem cells. Cell 161, 1187–1201 (2015).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  5. Macosko, E. Z. et al. Highly parallel genome-wide expression profiling of individual cells using nanoliter droplets. Cell 161, 1202–1214 (2015).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  6. Shalek, A. K. et al. Single-cell RNA-seq reveals dynamic paracrine control of cellular variation. Nature 510, 363–369 (2014).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  7. Rizvi, A. H. et al. Single-cell topological RNA-seq analysis reveals insights into cellular differentiation and development. Nat. Biotechnol. 35, 551–560 (2017).

    Article  PubMed  PubMed Central  CAS  Google Scholar 

  8. Gojo, J. et al. Single-cell RNA-seq reveals cellular hierarchies and impaired developmental trajectories in pediatric ependymoma. Cancer Cell 38, 44–59 (2020).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  9. Ståhl, P. L. et al. Visualization and analysis of gene expression in tissue sections by spatial transcriptomics. Science 353, 78–82 (2016).

    Article  PubMed  CAS  Google Scholar 

  10. Vickovic, S. et al. High-definition spatial transcriptomics for in situ tissue profiling. Nat. Methods 16, 987–990 (2019).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  11. Shah, S. et al. Dynamics and spatial genomics of the nascent transcriptome by intron seqFISH. Cell 174, 363–376 (2018).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  12. Eng, C.-H. L. et al. Transcriptome-scale super-resolved imaging in tissues by RNA seqFISH. Nature 568, 235–239 (2019).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  13. Chen, K. H., Boettiger, A. N., Moffitt, J. R., Wang, S. & Zhuang, X. Spatially resolved, highly multiplexed RNA profiling in single cells. Science 348, aaa6090 (2015).

    Article  PubMed  PubMed Central  CAS  Google Scholar 

  14. Moffitt, J. R. et al. High-throughput single-cell gene-expression profiling with multiplexed error-robust fluorescence in situ hybridization. Proc. Natl Acad. Sci. USA 113, 11046–11051 (2016).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  15. Wang, X. et al. Three-dimensional intact-tissue sequencing of single-cell transcriptional states. Science 361, eaat5691 (2018).

    Article  PubMed  PubMed Central  CAS  Google Scholar 

  16. Thompson, B. Canonical Correlation Analysis: Uses and Interpretation (Sage, 1984).

  17. Argelaguet, R. et al. Multi-Omics Factor Analysis—a framework for unsupervised integration of multi-omics data sets. Mol. Syst. Biol. 14, e8124 (2018).

    Article  PubMed  PubMed Central  CAS  Google Scholar 

  18. Argelaguet, R. et al. MOFA+: a statistical framework for comprehensive integration of multi-modal single-cell data. Genome Biol. 21, 111 (2020).

    Article  PubMed  PubMed Central  Google Scholar 

  19. Hinton, G.E. & Zemel, R.S. In: Advances in Neural Information Processing Systems 3–10 (MIT Press, 1994).

  20. Baldi, P. In: Proceedings of ICML Workshop on Unsupervised and Transfer Learning 37–49 (MLR Press, 2012).

  21. Chechik, G., Sharma, V., Shalit, U. & Bengio, S. Large scale online learning of image similarity through ranking. J.Mach. Learn. Res. 11, 1109–1135 (2010).

    Google Scholar 

  22. Hoffer, E. & Ailon, N. In: International Workshop on Similarity-Based Pattern Recognition 84–92 (Springer, 2015).

  23. Levine, J. H. et al. Data-driven phenotypic dissection of AML reveals progenitor-like cells that correlate with prognosis. Cell 162, 184–197 (2015).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  24. Hubert, L. & Arabie, P. Comparing partitions. J. Classif. 2, 193–218 (1985).

    Article  Google Scholar 

  25. Wang, B., Zhu, J., Pierson, E., Ramazzotti, D. & Batzoglou, S. Visualization and analysis of single-cell RNA-seq data by kernel-based similarity learning. Nat. Methods 14, 414–416 (2017).

    Article  CAS  PubMed  Google Scholar 

  26. Deng, Y., Bao, F., Dai, Q., Wu, L. F. & Altschuler, S. J. Scalable analysis of cell-type composition from single-cell transcriptomics using deep recurrent learning. Nat. Methods 16, 311–314 (2019).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  27. Qiu, P. Embracing the dropouts in single-cell RNA-seq analysis. Nat. Commun. 11, 1169 (2020).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  28. Yuan, G.-C. et al. Challenges and emerging directions in single-cell analysis. Genome Biol. 18, 84 (2017).

    Article  PubMed  PubMed Central  CAS  Google Scholar 

  29. Satija, R., Farrell, J. A., Gennert, D., Schier, A. F. & Regev, A. Spatial reconstruction of single-cell gene expression data. Nat. Biotechnol. 33, 495–502 (2015).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  30. Zhu, Q., Shah, S., Dries, R., Cai, L. & Yuan, G.-C. Identification of spatially associated subpopulations by combining scRNAseq and sequential fluorescence in situ hybridization data. Nat. Biotechnol. https://doi.org/10.1038/nbt.4260 (2018).

  31. Belgard, T. G. et al. A transcriptomic atlas of mouse neocortical layers. Neuron 71, 605–616 (2011).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  32. Tasic, B. et al. Adult mouse cortical cell taxonomy revealed by single cell transcriptomics. Nat. Neurosci. 19, 335–346 (2016).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  33. Szegedy, C., Vanhoucke, V., Ioffe, S., Shlens, J. & Wojna, Z. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition 2818–2826 (IEEE, 2016).

  34. Pierson, E. & Yau, C. ZIFA: dimensionality reduction for zero-inflated single-cell gene expression analysis. Genome Biol. 16, 241 (2015).

  35. Yao, Z. et al. A taxonomy of transcriptomic cell types across the isocortex and hippocampal formation. Cell 184, 3222–3241 (2021).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  36. Mbeunkui, F. & Johann, D. J. Cancer and the tumor microenvironment: a review of an essential relationship. Cancer Chemother. Pharmacol. 63, 571–582 (2009).

    Article  PubMed  Google Scholar 

  37. Sun, Y. et al. Treatment-induced damage to the tumor microenvironment promotes prostate cancer therapy resistance through WNT16B. Nat. Med. 18, 1359–1368 (2012).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  38. McGranahan, N. & Swanton, C. Clonal heterogeneity and tumor evolution: past, present, and the future. Cell 168, 613–628 (2017).

    Article  CAS  PubMed  Google Scholar 

  39. Moncada, R. et al. Integrating microarray-based spatial transcriptomics and single-cell RNA-seq reveals tissue architecture in pancreatic ductal adenocarcinomas. Nat. Biotechnol. 38, 333–342 (2020).

    Article  CAS  PubMed  Google Scholar 

  40. Yao, H. et al. Glypican-3 and KRT19 are markers associating with metastasis and poor prognosis of pancreatic ductal adenocarcinoma. Cancer Biomark. 17, 397–404 (2016).

    Article  CAS  PubMed  Google Scholar 

  41. Liu, X. et al. A new panel of pancreatic cancer biomarkers discovered using a mass spectrometry-based pipeline. Br. J. Cancer 117, 1846–1854 (2017).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  42. Roa-Peña, L. et al. Keratin 17 identifies the most lethal molecular subtype of pancreatic cancer. Sci. Rep. 9, 11239 (2019).

    Article  PubMed  PubMed Central  CAS  Google Scholar 

  43. Yang, C. et al. Evaluation of the diagnostic ability of laminin gene family for pancreatic ductal adenocarcinoma. Aging (Albany NY) 11, 3679–3703 (2019).

    Google Scholar 

  44. Van den Broeck, A., Vankelecom, H., Van Eijsden, R., Govaere, O. & Topal, B. Molecular markers associated with outcome and metastasis in human pancreatic cancer. J. Exp. Clin. Cancer Res. 31, 68 (2012).

    Article  PubMed  PubMed Central  CAS  Google Scholar 

  45. Fawkner-Corbett, D. et al. Spatiotemporal analysis of human intestinal development at single-cell resolution. Cell 184, 810–826 (2021).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  46. Maynard, K. R. et al. Transcriptome-scale spatial gene expression in the human dorsolateral prefrontal cortex. Nat. Neurosci. 24, 425–436 (2021).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  47. Zhao, E. et al. Spatial transcriptomics at subspot resolution with BayesSpace. Nat. Biotechnol. 39, 1375–1384 (2021).

  48. Hardy, J. A. & Higgins, G. A. Alzheimer’s disease: the amyloid cascade hypothesis. Science 256, 184–186 (1992).

    Article  CAS  PubMed  Google Scholar 

  49. Murphy, M. & Levine, H. III Alzheimer’s disease and the amyloid-β peptide. J. Alzheimers Dis. 19, 311–323 (2010).

    Article  PubMed  PubMed Central  CAS  Google Scholar 

  50. Chen, W.-T. et al. Spatial transcriptomics and in situ sequencing to study Alzheimer’s disease. Cell 182, 976–991(2020).

    Article  CAS  PubMed  Google Scholar 

  51. Stickels, R. R. et al. Highly sensitive spatial transcriptomics at near-cellular resolution with Slide-seqV2. Nat. Biotechnol. 39, 313–319 (2021).

    Article  CAS  PubMed  Google Scholar 

  52. Clark, S. J. et al. scNMT-seq enables joint profiling of chromatin accessibility DNA methylation and transcription in single cells. Nat. Commun. 9, 781 (2018).

  53. Dries, R. et al. Giotto: a toolbox for integrative analysis and visualization of spatial expression data. Genome Biol. 22, 78 (2021).

  54. Pham, D. et al. stLearn: integrating spatial location, tissue morphology and gene expression to find cell types, cell-cell interactions and spatial trajectories within undissociated tissues. Preprint at bioRxiv https://www.biorxiv.org/content/10.1101/2020.05.31.125658v1 (2020).

  55. Wolf, F. A., Angerer, P. & Theis, F. J. SCANPY: large-scale single-cell gene expression data analysis. Genome Biol. 19, 15 (2018).

    Article  PubMed  PubMed Central  Google Scholar 

  56. Eden, E., Navon, R., Steinfeld, I., Lipson, D. & Yakhini, Z. GOrilla: a tool for discovery and visualization of enriched GO terms in ranked gene lists. BMC Bioinformatics 10, 48 (2009).

  57. Subramanian, A. et al. Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proc. Natl Acad. Sci. USA 102, 15545–15550 (2005).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  58. Rousseeuw, P. J. Silhouettes: a graphical aid to the interpretation and validation of cluster analysis. J. Comput. Appl. Math. 20, 53–65 (1987).

    Article  Google Scholar 

Download references

Acknowledgements

We thank C.-H. L. Eng at Caltech for providing seqFISH+ image data; X. Wang at the Broad Institute and MIT for providing information on STARmap data analysis; R. Moncada at NYU for advice on PDAC data analysis; H. Koohy and A. Antanavicute from Oxford for providing full-resolution human intestine images; and O. Moindrot at Stanford for the open-source implementation of the triplet loss. We thank J. Bieber, H. Hammerlindl, L. Rao, X. Sun and other members of the Altschuler and Wu laboratories for constructive feedback. S.J.A. and L.F.W. gratefully acknowledge support from the UCSF Program for Breakthrough Biomedical Research, ProjectALS and the CZI NDNC Challenge Network. Q.D. was supported by the projects of NSFC (no. 62088102) and the MOST (no. 2020AA0105500). Y.D. was supported by the projects of NSFC (no. 61971020 and 62031001) and the MOST (no. 2020AAA0105502).

Author information

Author notes

  1. These authors contributed equally: Feng Bao, Yue Deng.

Authors and Affiliations

  1. Department of Pharmaceutical Chemistry, University of California, San Francisco, San Francisco, CA, USA

    Feng Bao, Susan Q. Shen, Steven J. Altschuler & Lani F. Wu

  2. School of Astronautics, Beihang University, Beijing, China

    Yue Deng & Bo Wang

  3. Institute of Artificial Intelligence, Beihang University, Beijing, China

    Yue Deng

  4. Department of Automation and Institute for Brain and Cognitive Science, Tsinghua University, Beijing, China

    Sen Wan & Qionghai Dai

  5. Department of Psychiatry, University of California, San Francisco, San Francisco, CA, USA

    Susan Q. Shen

  6. Beijing National Research Center for Information Science and Technology, Beijing, China

    Qionghai Dai

  7. Beijing Laboratory of Brain and Cognitive Intelligence and Beijing Key Laboratory of Multi-dimension & Multi-scale Computational Photography, Beijing, China

    Qionghai Dai

Authors
  1. Feng Bao
    View author publications

    You can also search for this author in PubMed Google Scholar

  2. Yue Deng
    View author publications

    You can also search for this author in PubMed Google Scholar

  3. Sen Wan
    View author publications

    You can also search for this author in PubMed Google Scholar

  4. Susan Q. Shen
    View author publications

    You can also search for this author in PubMed Google Scholar

  5. Bo Wang
    View author publications

    You can also search for this author in PubMed Google Scholar

  6. Qionghai Dai
    View author publications

    You can also search for this author in PubMed Google Scholar

  7. Steven J. Altschuler
    View author publications

    You can also search for this author in PubMed Google Scholar

  8. Lani F. Wu
    View author publications

    You can also search for this author in PubMed Google Scholar

Contributions

F.B., Y.D. and Q.D. developed the approach and conducted simulation experiments. F.B., Y.D., S.W., B.W., S.Q.S., S.J.A. and L.F.W. conducted experimental analyses on biological datasets. The manuscript was written by F.B., Y.D., S.Q.S., S.J.A. and L.F.W. All authors read and approved the manuscript.

Corresponding authors

Correspondence to Qionghai Dai, Steven J. Altschuler or Lani F. Wu.

Ethics declarations

Competing interests

S.J.A. and L.F.W. have consulting agreements with Nine Square Therapeutics and BAKX Therapeutics involving cash and/or equity compensation. All other authors declare no competing interests.

Peer review

Peer review information

Nature Biotechnology thanks Itai Yanai, Raphael Gottardo and the other, anonymous, reviewer(s) for their contribution to the peer review of this work.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Extended data

Extended Data Fig. 1 Overview and simulation studies of MUSE, related to Fig. 1.

Parameters used in simulation were listed in Supplementary Table 1. (a) Summary of data and analysis used in this work. (b) A flowchart of MUSE analysis pipeline. (c) Simulation design (Methods) to generate sample profiles with two modalities used for (d-s) below. (d) tSNE visualizations of latent representations from single- and combined-modality methods for randomly selected simulation experiments in Fig. 1c. Colors: ground-truth subpopulation labels in simulation. (e) Evaluation of combined methods in simulated data with different ground-truth cluster numbers. n = 1,000 (top) and 3,000 (bottom) samples were considered in simulations. (Note: for n = 1,000 and cluster number ≥30, each cluster may only contain a small number of samples.) (f) Evaluation of multi-modal methods in simulated data with Gaussian noise for increasing variance (σ). (g) Clustering accuracies for (i) analyses of concatenated modality features using various normalization approaches (Methods), and (ii) MUSE multi-modal analysis on matched or unmatched (randomly permuted sample order on one modality) data. ARIs were calculated based on n = 10 repeats. Boxplot: center line, median; box, interquartile range; whiskers, minimum–maximum range; same annotation also applies to other boxplots in this figure. (h) Example t-SNE visualization of MUSE subpopulations (indicated by shapes) and simple superimposition of single-modality clusters (indicated by colors) with simulation parameters chosen as in (f). (i) Simulation design using real morphological features from STARmap (Methods; dataset details were described in Fig. 3) and performances of multimodal methods (right n = 10). (j) Multimodal analysis on data with homogeneous features in one modality. Transcript profiles (left) were generated from a normal distribution while morphological features (middle) were simulated from known subpopulations as before. (k) Evaluation of clustering accuracy under different dimensions of joint latent representations (n = 10). (l) Clustering accuracy of MUSE while changing dimension of morphological features between 100 to 1,000 (n = 10). (m) Clustering accuracy of MUSE when fixing the latent representation of single modality (hx, hy) to different dimensions. ARIs were averaged on 10 repeats. Red underlines: parameters selected as default. (n) Effects of clustering methods on accuracies (n = 10). Cluster numbers for hierarchical and Kmeans methods were chosen using the elbow method with distortion score. (o) Run times for compared methods on simulated data; n = 1,000 cells. Note: for fair comparison, all methods were run under CPU mode. (p) Run times of MUSE on datasets with larger sample sizes using different clustering methods in label updating during training. (q) Accuracies and run times when fixing single modality labels (denoted as lx and ly in Methods) to the initial labels in training. Each dot represented one independent experiment. (r) Model structure of multi-modal autoencoder used in MUSE. (s) Performance evaluation of MUSE with different hyperparameter settings (n = 10): 1) weight of regularization term; 2) weight of supervision term; 3) learning rate; and 4) iteration intervals between cluster updating in training. Red underlines: parameters selected as default in MUSE package. (t) F-norms of selective matrices wx and wy to different true cluster numbers (left) in data and choices of regularization hyperparameter \(\lambda _{{{{\mathrm{regularization}}}}}\) (right); n = 10. (u) Clustering accuracies (left) and number of clusters (right) from PhenoGraph when change the hyperparameter of n_neighbor (n = 10).

Extended Data Fig. 2 Analysis of mouse cortex dataset from seqFISH+, related to Fig. 2.

(a) tSNE visualization of latent space from deep image features, overlaid with various cellular properties from CellProfiler. (b) Layer annotations of MUSE clusters based on layer gene markers. Spatial localization of cell clusters (first column) and marker expression abundances (second column) were shown. For each cluster, gene names with maximal overexpression levels were underlined. Boxplot: center line, median; box, interquartile range; whiskers, minimum–maximum range. (c) Comparison of discovered cortical layers by transcriptional or combined methods. 5 layers are shown. Squares with the same color and across multiple layers indicate the method discovered merged layers. Squares with no color indicate the method failed to discover the corresponding layer. (d) Subclustering analysis on transcript L2/3/4 cluster from Fig. 2c. Kmeans clustering were performed to divide L2/3/4 into two subclusters (middle). Spatial coherences with cortex layers were shown using cell density plots (right). (e) Comparisons of subpopulations identified by different clustering methods from multimodal features. In Kmeans, target cluster number (k) was set to the subpopulation size from MUSE analysis. (f) Shared up- and down-regulated glutamatergic marker genes between MUSE clusters and cell types from Allen Brain Atlas. Marker genes were obtained from recent Allen Brain Atlas publication; 36 markers were measured in both the seqFISH+ and Allen Brain datasets.

Extended Data Fig. 3 Comparison of methods on mouse cortex dataset from STARmap, related to Fig. 3.

(a) tSNE visualization of latent representations by different methods with pseudo-colors labeling cortex depth along x-coordinate (on right side). (b) Comparison of cell clusters on (top) numbers of identified clusters with or without significant spatial co-localization properties and (bottom) feature quality evaluation by cluster compactness in latent space using Silhouette coefficient. (c) Stability analysis of identified clusters to the choice of hyperparameter n_neighbor in PhenoGraph. Red circles: major differences in subpopulations compared with the result using default parameters (left panel) annotated with affected cortex layers. (d) Spatial mapping and annotations of clusters with significant spatial co-localization patterns. Significantly co-localized clusters are identified using spatial co-localization score with permutation test. Clusters are assigned to one layer with respect to the anatomic annotations by original paper (Methods). (e) tSNE visualization of MUSE clusters in MUSE latent space. All clusters were classified into ‘Refined’, ‘Reproduced’ or ‘Discovered’ types based on comparison with clusters identified from transcript-alone or morphological-alone analysis (corresponding to Fig. 3a). (f) 3D mapping of three types of MUSE clusters in the latent space of morphological features (top layer of each 3D plot), MUSE latent features (middle layer) or transcriptional features (bottom layer). Lines connect the same cells across the three spaces.

Extended Data Fig. 4 Application of MUSE to a multimodal pancreatic ductal adenocarcinoma (PDAC) dataset, related to Fig. 4.

(a) tSNE visualizations of latent representations and identified clusters by transcripts-alone (left), H&E image-alone (middle) and MUSE (right) analyses (corresponding to Fig. 4a). (b) Manual histological annotations (colored lines) provided in original publications overlaid with regional clusters (colored circles) from image analysis. Highlighted regions show the morphological differences. (c-d) Analysis of single-cell RNA-seq data from the same PDAC tissue. tSNE visualization with cell type annotations (c) and signature gene expressions of two cancer clones (d). Cell type annotations from original publication. (e) Subclustering analysis of transcript cancer region using Kmeans method and comparisons of clone signature expressions between transcript subclusters and MUSE cancer regions. Boxplot: center line, median; box, interquartile range; whiskers, minimum–maximum range. n = 44 for subcluster 1 and n = 71 for subcluster 2. (f) Spatial expression maps of overexpressed genes in cancer regions (top) or pancreatic tissues (bottom) through differential expression analysis between pancreatic tissue regions and cancer regions characterized by MUSE (Methods) (g) Cluster separateness of tissue image spots with different size. We segmented image tiles with different pixel sizes and input them into Inception-v3 to learn deep features. Then we performed clustering on features and used Silhouette score to quantify the separateness of clusters. Red arrow indicates the chosen region size.

Extended Data Fig. 5 Application of MUSE to a Visium human intestine dataset, related to Fig. 5.

(a-b) tSNE visualizations of latent representations (a) and spatial plots (b) of identified clusters by transcripts-alone (left), H&E image-alone (middle) and MUSE (right) analyses. (c) Selected regions with various morphological patterns in the tissue. (d) Enhanced spatial maps of subpopulations from BayesSpace (left) or BayesSpace + MUSE (right). Details of the analysis were provided in Methods. (e) Selected zoom-in region examples with marker gene expressions or morphological patterns (top) and subpopulations defined from BayesSpace (middle) and MUSE (bottom) for four analyzed cell types in Fig. 5.

Extended Data Fig. 6 Application of MUSE to a multimodal Alzheimer’s disease dataset, related to Fig. 6.

(a) A summary of samples collected in the Alzheimer’s disease dataset. (b) tSNE was fitted on MUSE deep embeddings and each spot was colored by the Aβ index (defined by standard deviation of intensity in the previous study). (c) Visualization of deep embeddings of Aβ spots in the same ages. Color annotations as in (a). (d) Proportion of samples from all 4 timepoints in each MUSE cluster, related to Fig. 6e.

Supplementary information

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Bao, F., Deng, Y., Wan, S. et al. Integrative spatial analysis of cell morphologies and transcriptional states with MUSE. Nat Biotechnol 40, 1200–1209 (2022). https://doi.org/10.1038/s41587-022-01251-z

Download citation

  • Received:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1038/s41587-022-01251-z

This article is cited by

Search

Advanced search

Quick links

Nature Briefing

Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

Get the most important science stories of the day, free in your inbox. Sign up for Nature Briefing