A1325
Title: Spatial information integration for analyzing cellular drug responses
Authors: Mahiro Yamamoto - Doshisha University (Japan) [presenting]
Masaaki Okabe - Nagoya University (Japan)
Hiroshi Yadohisa - Doshisha University (Japan)
Abstract: Single-cell RNA sequencing (scRNA-seq) provides high-resolution gene expression profiles but lacks spatial context, limiting tissue-level analysis. Spatial transcriptomics (ST) preserves spatial information but cannot achieve single-cell resolution. While optimal transport methods can integrate ST spatial information with scRNA-seq data, drug responses analysis (perturbation analysis) remains challenging, where post-perturbation ST data is not readily obtainable. The aim is to propose a novel approach using neural optimal transport to propagate spatial information from pre-perturbation spatially-annotated scRNA-seq data to post-perturbation scRNA-seq data. The method models cellular state transitions by learning transport mappings between gene expression distributions via neural networks, enabling correspondence between unpaired pre- and post-perturbation cells. This framework addresses the critical gap in spatially-aware perturbation analysis by leveraging pre-perturbation ST-scRNA-seq integration as a foundation for spatial context propagation. The method enables comprehensive analysis of spatially-resolved gene expression patterns in response to drug responses. The contribution enables spatial context-aware analysis of drug responses where direct spatial measurement is technically challenging, opening new possibilities for understanding spatially dependent cellular responses to various experimental drug responses.
