A novel multislice framework for precision 3D spatial domain reconstruction and disease pathology analysis

Area of Science:

• Genomics
• Computational Biology
• Pathology

Background:

• Spatial transcriptomics (ST) technologies offer insights into tissue organization and function.
• Current ST methods often lack comprehensive capture of multilayered biological heterogeneity due to single-modality or resolution limitations.

Purpose of the Study:

• To introduce STMSC, a multislice joint analysis framework with precorrection for precise identification of complex spatial domains.
• To enhance 3D reconstruction accuracy by incorporating hematoxylin and eosin (H&E) imaging data.
• To provide deeper insights into complex cancer environments and pathological tissue structures.

Main Methods:

• Developed STMSC, a multislice joint analysis framework with a precorrection mechanism.
• Incorporated hematoxylin and eosin (H&E) imaging data for enhanced slice alignment in 3D reconstruction.
• Utilized a graph attention autoencoder with precorrection to balance biological information across different levels.

Main Results:

• STMSC enables precise identification of complex spatial domains and reconstruction of fine-grained cellular landscapes.
• The framework accurately reconstructs 3D tissue structures by analyzing consecutive slices and pathological data.
• Achieved deeper insights into complex cancer environments, capturing intra- and interstage heterogeneity.

Conclusions:

• STMSC advances disease pathology insights by enabling precise 3D reconstruction and analysis of spatial transcriptomics data.
• The framework effectively deconstructs microenvironments and emphasizes collective cellular behavior for defining spatial domains.
• STMSC offers novel insights into the complexity of pathological tissue structures and cancer development.