SINTER3D: continuous 3D reconstruction of spatial transcriptomics via implicit neural representations.

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Spatial transcriptomics provides gene expression data with spatial context.
• Current 3D reconstruction methods struggle with large gaps between tissue sections and independent gene interpolation.
• Accurate 3D molecular structure reconstruction is crucial for understanding tissue organization and function.

Purpose of the Study:

• To develop an advanced framework for 3D spatial transcriptomics reconstruction.
• To overcome limitations of existing methods in interpolating gene expression data across large inter-section gaps.
• To enable more accurate and comprehensive 3D molecular mapping of tissues.

Main Methods:

• Developed SINTER3D, an implicit neural representation-based framework.
• Modeled gene expression as continuous functions of 3D coordinates for joint interpolation of multiple genes.
• Applied the framework to diverse biological datasets including brain, heart, embryo, and cancer tissues.

Main Results:

• SINTER3D demonstrated superior performance compared to existing methods across multiple datasets.
• The framework successfully reconstructed biologically meaningful 3D molecular structures.
• Enabled accurate virtual section generation, spatial-domain identification, and cell-type deconvolution.

Conclusions:

• SINTER3D offers a significant advancement in 3D spatial transcriptomics reconstruction.
• The implicit neural representation approach effectively addresses challenges in data interpolation and 3D modeling.
• This method enhances the ability to analyze and interpret complex 3D molecular landscapes in various biological contexts.