Mapping the topography of spatial gene expression with interpretable deep learning

Uthsav Chitra; Brian J. Arnold; Hirak Sarkar; Kohei Sanno; Cong Ma; Sereno Lopez-Darwin; Benjamin J. Raphael

doi:10.1038/s41592-024-02503-3

Mapping the topography of spatial gene expression with interpretable deep learning

Uthsav Chitra, Brian J. Arnold, Hirak Sarkar, Kohei Sanno, Cong Ma, Sereno Lopez-Darwin, Benjamin J. Raphael

Research output: Contribution to journal › Article › peer-review

Abstract

Spatially resolved transcriptomics technologies provide high-throughput measurements of gene expression in a tissue slice, but the sparsity of these data complicates analysis of spatial gene expression patterns. We address this issue by deriving a topographic map of a tissue slice—analogous to a map of elevation in a landscape—using a quantity called the isodepth. Contours of constant isodepths enclose domains with distinct cell type composition, while gradients of the isodepth indicate spatial directions of maximum change in expression. We develop GASTON (gradient analysis of spatial transcriptomics organization with neural networks), an unsupervised and interpretable deep learning algorithm that simultaneously learns the isodepth, spatial gradients and piecewise linear expression functions that model both continuous gradients and discontinuous variation in gene expression. We show that GASTON accurately identifies spatial domains and marker genes across several tissues, gradients of neuronal differentiation and firing in the brain, and gradients of metabolism and immune activity in the tumor microenvironment.

Original language	American English
Article number	8353
Pages (from-to)	298-309
Number of pages	12
Journal	Nature Methods
Volume	22
Issue number	2
DOIs	https://doi.org/10.1038/s41592-024-02503-3
State	Published - Feb 2025
Externally published	Yes

ASJC Scopus subject areas

Biotechnology
Biochemistry
Molecular Biology
Cell Biology

Access to Document

10.1038/s41592-024-02503-3

Cite this

@article{083b76016c06499f9950b49f4dc6f346,

title = "Mapping the topography of spatial gene expression with interpretable deep learning",

abstract = "Spatially resolved transcriptomics technologies provide high-throughput measurements of gene expression in a tissue slice, but the sparsity of these data complicates analysis of spatial gene expression patterns. We address this issue by deriving a topographic map of a tissue slice—analogous to a map of elevation in a landscape—using a quantity called the isodepth. Contours of constant isodepths enclose domains with distinct cell type composition, while gradients of the isodepth indicate spatial directions of maximum change in expression. We develop GASTON (gradient analysis of spatial transcriptomics organization with neural networks), an unsupervised and interpretable deep learning algorithm that simultaneously learns the isodepth, spatial gradients and piecewise linear expression functions that model both continuous gradients and discontinuous variation in gene expression. We show that GASTON accurately identifies spatial domains and marker genes across several tissues, gradients of neuronal differentiation and firing in the brain, and gradients of metabolism and immune activity in the tumor microenvironment.",

author = "Uthsav Chitra and Arnold, {Brian J.} and Hirak Sarkar and Kohei Sanno and Cong Ma and Sereno Lopez-Darwin and Raphael, {Benjamin J.}",

note = "Publisher Copyright: {\textcopyright} The Author(s), under exclusive licence to Springer Nature America, Inc. 2025.",

year = "2025",

month = feb,

doi = "10.1038/s41592-024-02503-3",

language = "American English",

volume = "22",

pages = "298--309",

journal = "Nature Methods",

issn = "1548-7091",

publisher = "Lippincott Williams and Wilkins Ltd.",

number = "2",

}

TY - JOUR

T1 - Mapping the topography of spatial gene expression with interpretable deep learning

AU - Chitra, Uthsav

AU - Arnold, Brian J.

AU - Sarkar, Hirak

AU - Sanno, Kohei

AU - Ma, Cong

AU - Lopez-Darwin, Sereno

AU - Raphael, Benjamin J.

N1 - Publisher Copyright: © The Author(s), under exclusive licence to Springer Nature America, Inc. 2025.

PY - 2025/2

Y1 - 2025/2

N2 - Spatially resolved transcriptomics technologies provide high-throughput measurements of gene expression in a tissue slice, but the sparsity of these data complicates analysis of spatial gene expression patterns. We address this issue by deriving a topographic map of a tissue slice—analogous to a map of elevation in a landscape—using a quantity called the isodepth. Contours of constant isodepths enclose domains with distinct cell type composition, while gradients of the isodepth indicate spatial directions of maximum change in expression. We develop GASTON (gradient analysis of spatial transcriptomics organization with neural networks), an unsupervised and interpretable deep learning algorithm that simultaneously learns the isodepth, spatial gradients and piecewise linear expression functions that model both continuous gradients and discontinuous variation in gene expression. We show that GASTON accurately identifies spatial domains and marker genes across several tissues, gradients of neuronal differentiation and firing in the brain, and gradients of metabolism and immune activity in the tumor microenvironment.

AB - Spatially resolved transcriptomics technologies provide high-throughput measurements of gene expression in a tissue slice, but the sparsity of these data complicates analysis of spatial gene expression patterns. We address this issue by deriving a topographic map of a tissue slice—analogous to a map of elevation in a landscape—using a quantity called the isodepth. Contours of constant isodepths enclose domains with distinct cell type composition, while gradients of the isodepth indicate spatial directions of maximum change in expression. We develop GASTON (gradient analysis of spatial transcriptomics organization with neural networks), an unsupervised and interpretable deep learning algorithm that simultaneously learns the isodepth, spatial gradients and piecewise linear expression functions that model both continuous gradients and discontinuous variation in gene expression. We show that GASTON accurately identifies spatial domains and marker genes across several tissues, gradients of neuronal differentiation and firing in the brain, and gradients of metabolism and immune activity in the tumor microenvironment.

UR - http://www.scopus.com/inward/record.url?scp=85217246315&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85217246315&partnerID=8YFLogxK

U2 - 10.1038/s41592-024-02503-3

DO - 10.1038/s41592-024-02503-3

M3 - Article

C2 - 39849132

SN - 1548-7091

VL - 22

SP - 298

EP - 309

JO - Nature Methods

JF - Nature Methods

IS - 2

M1 - 8353

ER -

Mapping the topography of spatial gene expression with interpretable deep learning

Abstract

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this