Spatial transcriptomics prediction from histology jointly through Transformer and graph neural networks DOI
Yuansong Zeng,

Zhuoyi Wei,

Weijiang Yu

et al.

Briefings in Bioinformatics, Journal Year: 2022, Volume and Issue: 23(5)

Published: July 17, 2022

Abstract The rapid development of spatial transcriptomics allows the measurement RNA abundance at a high resolution, making it possible to simultaneously profile gene expression, locations cells or spots, and corresponding hematoxylin eosin-stained histology images. It turns promising predict expression from images that are relatively easy cheap obtain. For this purpose, several methods devised, but they have not fully captured internal relations 2D vision features dependency between spots. Here, we developed Hist2ST, deep learning-based model RNA-seq Around each sequenced spot, image is cropped into an patch fed convolutional module extract features. Meanwhile, with whole neighbored patches through Transformer graph neural network modules, respectively. These learned then used by following zero-inflated negative binomial distribution. To alleviate impact small data, self-distillation mechanism employed for efficient learning model. By comprehensive tests on cancer normal datasets, Hist2ST was shown outperform existing in terms both prediction region identification. Further pathway analyses indicated our could reserve biological information. Thus, enables generating data elucidating molecular signatures tissues.

Language: Английский

Spatial components of molecular tissue biology DOI
Giovanni Palla, David S. Fischer, Aviv Regev

et al.

Nature Biotechnology, Journal Year: 2022, Volume and Issue: 40(3), P. 308 - 318

Published: Feb. 7, 2022

Language: Английский

Citations

237
Spatially informed clustering, integration, and deconvolution of spatial transcriptomics with GraphST DOI Creative Commons
Yahui Long, Kok Siong Ang, Mengwei Li

et al.

Nature Communications, Journal Year: 2023, Volume and Issue: 14(1)

Published: March 1, 2023

Abstract Spatial transcriptomics technologies generate gene expression profiles with spatial context, requiring spatially informed analysis tools for three key tasks, clustering, multisample integration, and cell-type deconvolution. We present GraphST, a graph self-supervised contrastive learning method that fully exploits data to outperform existing methods. It combines neural networks learn informative discriminative spot representations by minimizing the embedding distance between adjacent spots vice versa. demonstrated GraphST on multiple tissue types technology platforms. achieved 10% higher clustering accuracy better delineated fine-grained structures in brain embryo tissues. is also only can jointly analyze slices vertical or horizontal integration while correcting batch effects. Lastly, superior deconvolution capture niches like lymph node germinal centers exhausted tumor infiltrating T cells breast tissue.

Language: Английский

Citations

233
Screening cell–cell communication in spatial transcriptomics via collective optimal transport DOI Creative Commons
Zixuan Cang, Yanxiang Zhao, Axel A. Almet

et al.

Nature Methods, Journal Year: 2023, Volume and Issue: 20(2), P. 218 - 228

Published: Jan. 23, 2023

Abstract Spatial transcriptomic technologies and spatially annotated single-cell RNA sequencing datasets provide unprecedented opportunities to dissect cell–cell communication (CCC). However, incorporation of the spatial information complex biochemical processes required in reconstruction CCC remains a major challenge. Here, we present COMMOT (COMMunication analysis by Optimal Transport) infer transcriptomics, which accounts for competition between different ligand receptor species as well distances cells. A collective optimal transport method is developed handle molecular interactions constraints. Furthermore, introduce downstream tools signaling directionality genes regulated using machine learning models. We apply simulation data eight acquired with five show its effectiveness robustness identifying varying resolutions gene coverages. Finally, identifies new CCCs during skin morphogenesis case study human epidermal development.

Language: Английский

Citations

200
Advances in spatial transcriptomic data analysis DOI Creative Commons
Ruben Dries,

Jiaji Chen,

Natalie Del Rossi

et al.

Genome Research, Journal Year: 2021, Volume and Issue: 31(10), P. 1706 - 1718

Published: Oct. 1, 2021

Spatial transcriptomics is a rapidly growing field that promises to comprehensively characterize tissue organization and architecture at the single-cell or subcellular resolution. Such information provides solid foundation for mechanistic understanding of many biological processes in both health disease cannot be obtained by using traditional technologies. The development computational methods plays important roles extracting signals from raw data. Various approaches have been developed overcome technology-specific limitations such as spatial resolution, gene coverage, sensitivity, technical biases. Downstream analysis tools formulate cell–cell communications quantifiable properties, provide algorithms derive properties. Integrative pipelines further assemble multiple one package, allowing biologists conveniently analyze data beginning end. In this review, we summarize state art transcriptomic pipelines, discuss how they operate on different technological platforms.

Language: Английский

Citations

172
The landscape of cell–cell communication through single-cell transcriptomics DOI Creative Commons
Axel A. Almet, Zixuan Cang, Suoqin Jin

et al.

Current Opinion in Systems Biology, Journal Year: 2021, Volume and Issue: 26, P. 12 - 23

Published: March 26, 2021

Cell-cell communication is a fundamental process that shapes biological tissue. Historically, studies of cell-cell have been feasible for one or two cell types and few genes. With the emergence single-cell transcriptomics, we are now able to examine genetic profiles individual cells at unprecedented scale depth. The availability such data presents an exciting opportunity construct more comprehensive description communication. This review discusses recent explosion methods developed infer from non-spatial spatial promising technologies which complementary strengths limitations. We propose several avenues propel this rapidly expanding field forward in meaningful ways.

Language: Английский

Citations

147
DeepST: identifying spatial domains in spatial transcriptomics by deep learning DOI Creative Commons

Chang Xu,

Xiyun Jin, Songren Wei

et al.

Nucleic Acids Research, Journal Year: 2022, Volume and Issue: 50(22), P. e131 - e131

Published: Oct. 4, 2022

Recent advances in spatial transcriptomics (ST) have brought unprecedented opportunities to understand tissue organization and function context. However, it is still challenging precisely dissect domains with similar gene expression histology situ. Here, we present DeepST, an accurate universal deep learning framework identify domains, which performs better than the existing state-of-the-art methods on benchmarking datasets of human dorsolateral prefrontal cortex. Further testing a breast cancer ST dataset, showed that DeepST can at finer scale. Moreover, achieve not only effective batch integration data generated from multiple batches or different technologies, but also expandable capabilities for processing other omics data. Together, our results demonstrate has exceptional capacity identifying making desirable tool gain novel insights studies.

Language: Английский

Citations

138
Spatially aware dimension reduction for spatial transcriptomics DOI Creative Commons
Lulu Shang, Xiang Zhou

Nature Communications, Journal Year: 2022, Volume and Issue: 13(1)

Published: Nov. 23, 2022

Abstract Spatial transcriptomics are a collection of genomic technologies that have enabled transcriptomic profiling on tissues with spatial localization information. Analyzing data is computationally challenging, as the collected from various often noisy and display substantial correlation across tissue locations. Here, we develop spatially-aware dimension reduction method, SpatialPCA, can extract low dimensional representation biological signal preserved structure, thus unlocking many existing computational tools previously developed in single-cell RNAseq studies for tailored analysis transcriptomics. We illustrate benefits SpatialPCA domain detection explores its utility trajectory inference high-resolution map construction. In real applications, identifies key molecular immunological signatures detected tumor surrounding microenvironment, including tertiary lymphoid structure shapes gradual transition during tumorigenesis metastasis. addition, detects past neuronal developmental history underlies current landscape locations cortex.

Language: Английский

Citations

136
Unsupervised spatially embedded deep representation of spatial transcriptomics DOI Creative Commons
Hang Xu, Huazhu Fu, Yahui Long

et al.

Genome Medicine, Journal Year: 2024, Volume and Issue: 16(1)

Published: Jan. 12, 2024

Abstract Optimal integration of transcriptomics data and associated spatial information is essential towards fully exploiting to dissect tissue heterogeneity map out inter-cellular communications. We present SEDR, which uses a deep autoencoder coupled with masked self-supervised learning mechanism construct low-dimensional latent representation gene expression, then simultaneously embedded the corresponding through variational graph autoencoder. SEDR achieved higher clustering performance on manually annotated 10 × Visium datasets better scalability high-resolution than existing methods. Additionally, we show SEDR’s ability impute denoise expression (URL: https://github.com/JinmiaoChenLab/SEDR/ ).

Language: Английский

Citations

126
Statistical and machine learning methods for spatially resolved transcriptomics data analysis DOI Creative Commons
Zexian Zeng, Yawei Li, Yiming Li

et al.

Genome biology, Journal Year: 2022, Volume and Issue: 23(1)

Published: March 25, 2022

Abstract The recent advancement in spatial transcriptomics technology has enabled multiplexed profiling of cellular transcriptomes and locations. As the capacity efficiency experimental technologies continue to improve, there is an emerging need for development analytical approaches. Furthermore, with continuous evolution sequencing protocols, underlying assumptions current methods be re-evaluated adjusted harness increasing data complexity. To motivate aid future model development, we herein review statistical machine learning transcriptomics, summarize useful resources, highlight challenges opportunities ahead.

Language: Английский

Citations

125
Cell clustering for spatial transcriptomics data with graph neural networks DOI
Jiachen Li, Siheng Chen, Xiaoyong Pan

et al.

Nature Computational Science, Journal Year: 2022, Volume and Issue: 2(6), P. 399 - 408

Published: June 27, 2022

Language: Английский

Citations

119