Fault-tolerant 3D reconstruction from 2D spatial proteomics sections.

Area of Science:

• Biotechnology
• Computational Biology
• Molecular Imaging

Background:

• Reconstructing 3D molecular volumes from 2D tissue sections is challenging due to data limitations.
• Marker dropout and tissue loss in sections hinder accurate 3D reconstruction.
• Current methods struggle with sparsely sampled or damaged 2D section data.

Purpose of the Study:

• To develop a computational pipeline for robust 3D molecular volume reconstruction from imperfect 2D tissue sections.
• To enable 3D spatial proteomics at single-cell resolution despite data sparsity and damage.
• To facilitate the construction of 3D atlases from practical, albeit imperfect, section data.

Main Methods:

• Developed 3D-Omics-Flow, a generative pipeline for 3D molecular volume reconstruction.
• The pipeline jointly repairs damaged 2D sections and interpolates missing data.
• Applied the method to diverse datasets across health and disease states.

Main Results:

• Successfully reconstructed 3D molecular volumes from sparsely sampled and damaged 2D sections.
• Demonstrated the pipeline's ability to handle marker dropout and tissue loss.
• Enabled 3D spatial proteomics in practical sampling regimes.

Conclusions:

• 3D-Omics-Flow overcomes limitations in 3D molecular volume reconstruction from imperfect 2D sections.
• The pipeline facilitates the creation of comprehensive 3D spatial atlases.
• This advancement supports downstream analyses in various biological contexts.