Global Context-aware Representation Learning for Spatially Resolved Transcriptomics
Keywords: Spatially Resolved Transcriptomics, Self-Supervised Learning
Abstract: Spatially Resolved Transcriptomics (SRT) is a cutting-edge technique that captures the spatial context of cells within tissues, enabling the study of complex biological networks. Recently, graph-based deep learning has been utilized in identifying meaningful spatial domains by leveraging both gene expression and spatial information. However, these approaches fall short in obtaining qualified spot representations, particularly for those located around the boundary of cell type clusters, as they heavily emphasize spatially local spots that have minimal feature differences from an anchor node. To address this limitation, we propose a novel framework, Spotscape, which introduces the Similarity Telescope module designed to learn spot representations by capturing the global relationships among multiple spots. Additionally, to address the challenges that arise when integrating multiple slices from heterogeneous sources, we propose a similarity scaling strategy that explicitly regulates the distances between intra- and inter-slice spots to ensure they remain nearly the same. Extensive experiments demonstrate the superiority of Spotscape in various downstream tasks, including spatial domain identification, multi-slice integration, and alignment tasks, compared to baseline methods. Our code is available at the following link: https://anonymous.4open.science/r/Spotscape-E312/
Primary Area: applications to physical sciences (physics, chemistry, biology, etc.)
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 14092
Loading