SpatialMETA

SpatialMETA: Integrating Cross-Sample and Cross-Modal Data for Spatial Transcriptomics and Metabolomics

Status: Successfully concluded and accepted for publication in Nature Communications (2025)

Role: Research Assistant, Co-author

Institution: Prof. Wanlu Liu Lab, ZJU-UoE Institute, Zhejiang University

Duration: November 2024 – Spring 2025

Project Overview

SpatialMETA is an innovative algorithm development project that uses conditional variation autoencoder (cVAE) to enable the integration of spatial transcriptomics and spatial metabolomics data. This work addresses a critical challenge in spatial multi-omics analysis by providing a robust computational framework for cross-modal data integration.

My Contributions

• Benchmark Development: Designed and implemented comprehensive benchmark datasets and evaluation metrics specifically for the multimodal spatial integration task
• Method Validation: Conducted extensive validation studies to assess the performance of the integration algorithm
• Performance Analysis: Developed quantitative metrics to evaluate the quality of cross-modal spatial data integration

Technical Approach

The project leverages advanced machine learning techniques, specifically conditional variational autoencoders, to:

• Learn shared representations between spatial transcriptomics and metabolomics data
• Preserve spatial relationships during cross-modal integration
• Enable downstream analysis of integrated spatial multi-omics datasets

Impact

This work provides the scientific community with a powerful tool for spatial multi-omics analysis, enabling researchers to gain deeper insights into tissue architecture and cellular function through integrated analysis of transcriptomic and metabolomic data.

Publication: Tian R†, Xue Z†, Chen Y, Qi Y, Zhang J, Yuan J, Ruan D, Lin J, Liu J, Wang D, Youqiong Y, Liu W*. Integrating Cross-Sample and Cross-Modal Data for Spatial Transcriptomics and Metabolomics with SpatialMETA. *Nature Communications* (Accepted), 2025.