TY - GEN
T1 - Joint Imputation and Deconvolution of Gene Expression Across Spatial Transcriptomics Platforms
AU - Zheng, Hongyu
AU - Sarkar, Hirak
AU - Raphael, Benjamin J.
N1 - Publisher Copyright:
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.
PY - 2025
Y1 - 2025
N2 - Spatially resolved transcriptomics (SRT) technologies measure gene expression across thousands of spatial locations in a tissue, with different technologies having varying resolution, gene coverage, and sequencing depth. Integration of multiple SRT technologies can overcome the limitations of individual technologies; for example enabling gene expression imputation in spatial technologies with limited gene coverage (e.g. 10x Genomics Xenium) or deconvolution of cells in technologies with low spatial resolution (e.g. 10x Genomics Visium). We introduce Spatial Integration for Imputation and Deconvolution (SIID), a joint non-negative factorization model that aligns and integrates paired SRT datasets that profile nearby tissue slices with different SRT technologies. We show that SIID outperforms existing tools in reconstructing cell-type assignments, recovering gene expression, and imputing missing data on simulated data and 10x Genomics Xenium-Visium paired datasets from breast and colon cancer.
AB - Spatially resolved transcriptomics (SRT) technologies measure gene expression across thousands of spatial locations in a tissue, with different technologies having varying resolution, gene coverage, and sequencing depth. Integration of multiple SRT technologies can overcome the limitations of individual technologies; for example enabling gene expression imputation in spatial technologies with limited gene coverage (e.g. 10x Genomics Xenium) or deconvolution of cells in technologies with low spatial resolution (e.g. 10x Genomics Visium). We introduce Spatial Integration for Imputation and Deconvolution (SIID), a joint non-negative factorization model that aligns and integrates paired SRT datasets that profile nearby tissue slices with different SRT technologies. We show that SIID outperforms existing tools in reconstructing cell-type assignments, recovering gene expression, and imputing missing data on simulated data and 10x Genomics Xenium-Visium paired datasets from breast and colon cancer.
KW - deconvolution
KW - imputation
KW - non-negative matrix factorization
KW - Spatial transcriptomics
UR - http://www.scopus.com/inward/record.url?scp=105004252776&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=105004252776&partnerID=8YFLogxK
U2 - 10.1007/978-3-031-90252-9_39
DO - 10.1007/978-3-031-90252-9_39
M3 - Conference contribution
AN - SCOPUS:105004252776
SN - 9783031902512
T3 - Lecture Notes in Computer Science
SP - 358
EP - 361
BT - Research in Computational Molecular Biology - 29th International Conference, RECOMB 2025, Proceedings
A2 - Sankararaman, Sriram
PB - Springer Science and Business Media Deutschland GmbH
T2 - 29th International Conference on Research in Computational Molecular Biology, RECOMB 2025
Y2 - 26 April 2025 through 29 April 2025
ER -