Unified integration of spatial transcriptomics across platforms with LLOKI.

Area of Science:

• Genomics and Bioinformatics
• Computational Biology
• Molecular Biology

Background:

• Spatial transcriptomics (ST) offers insights into tissue architecture and cellular interactions.
• Integrating ST data across platforms is challenging due to varying gene panels, data sparsity, and technical variability.

Purpose of the Study:

• To introduce LLOKI, a novel framework for integrating imaging-based ST data from diverse platforms.
• To enable cross-platform ST analysis without requiring shared gene panels.

Main Methods:

• LLOKI employs feature alignment across technologies and batch alignment across datasets.
• Optimal transport-guided feature propagation and graph-based imputation match ST data to scRNA-seq references.
• Single-cell foundation models like scGPT generate unified features, followed by batch alignment to refine embeddings.

Main Results:

• LLOKI successfully integrates ST data from five different mouse brain technologies, outperforming existing methods.
• The framework enables effective cross-technology spatial gene program identification and tissue slice alignment.
• Application to ovarian cancer datasets identified an integrated gene program of tumor-infiltrating T cells.

Conclusions:

• LLOKI provides a robust solution for cross-platform spatial transcriptomics studies.
• The framework has the potential to scale to large atlas datasets for deeper biological insights.
• LLOKI facilitates comprehensive analysis of cellular organization and tissue microenvironments.