论文阅读记录

stMMR

• 标题：stMMR: accurate and robust saptial domain identification from spatially resolved transcriptomics with multimodal feature representation
• 作者：Shandong University
• 发表会议/期刊：GigaScience
• 年份：2024
• 链接：stMMR

主要内容简介

• Motivation:
  融合多模态困难，多模态数据间具有显著异质性，而且在数据尺度和分辨率上存在差异

方法与创新点

• 总体框架：
  3 steps: multimodal feature embedding, feature fusion, and feature reconstruction
  

• 方法概述：

  1. Multimodal feature embedding

     • gene expression: $ G \in \mathbb{R}^{n \times p} $ (n spots, p genes)

     • histological image: $ H \in \mathbb{R}^{n \times m} $ (m image features)

     • spatial location: 如图构建无向加权图
       

     • 2层GCN编码器进行图像特征H 和 基因表达特征G 的消息传递与聚合($E^0=H, E^0=G$,输出$E_H和E_G$)：
       

  2. Feature fusion

     • 单模态内($E_H和E_G$)spots之间的全局关系(a normalized attention module),输出$E_{AH}和E_{AG}$
       

     • 全连接降维，全连接融合，Loss，三部分整合
       
       

  3. Feature reconstruction

     • ZINB模型重构基因表达
       
     • MSE重构图像特征
       

  4. Objective function
     

Results

1. stMMR enhances detection of stratified architectural patterns in human dorsolateral prefrontal cortex tissue
2. stMMR enhances spatial gene expression profiling and structural characterization
3. stMMR deciphers evolving cell lineage structures in the chicken heart ST dataset
4. stMMR accurately identifies tumor region in human breast cancer
5. stMMR dissects cell-type differences in a lung cancer SRT dataset based on NanoString technology

完结撒花