GMSA: A Graph Matching and Point Cloud Registration-Based Method for Spatial Transcriptomics Data Alignment

Area of Science:

• Computational Biology
• Bioinformatics
• Genomics

Background:

• Spatial transcriptomics (ST) enables gene expression analysis within tissue context.
• Accurate alignment of multiple ST slices is crucial for 3D reconstruction.
• Complex tissue deformations and heterogeneity pose significant challenges to current alignment methods.

Purpose of the Study:

• To develop a robust and accurate framework for aligning multiple spatial transcriptomics slices.
• To improve the reconstruction of three-dimensional biological structures from ST data.
• To address limitations of existing methods in handling complex tissue deformations.

Main Methods:

• Proposed GMSA (Graph Matching and Spatial Alignment), a synergistic framework integrating graph matching with point cloud registration.
• Utilized subgraph matching based on gene expression and spatial topology for robust correspondence identification.
• Employed Iterative Closest Point (ICP) and Nonrigid Iterative Closest Point (NICP) for refined alignment.

Main Results:

• GMSA demonstrated superior alignment accuracy on benchmark datasets including DLPFC, STARmap, and MERFISH.
• The nonrigid strategy of GMSA successfully resolved complex structural distortions in MERFISH data.
• GMSA maintained stable gene expression distributions across aligned slices, outperforming state-of-the-art methods.

Conclusions:

• GMSA offers a flexible and precise solution for multimodal spatial transcriptomics integration.
• The framework effectively addresses challenges in aligning heterogeneous and deformed tissue slices.
• GMSA advances the capability for accurate 3D reconstruction and analysis in spatial transcriptomics.