NicheTrans: spatial-aware cross-omics translation.

Area of Science:

• Computational Biology
• Systems Biology
• Genomics

Background:

• Spatial multiomics provides valuable insights into biological systems but faces adoption barriers due to technical complexity and accessibility issues.
• Existing single-cell translation methods often lack the ability to incorporate crucial cellular microenvironment information.
• There is a need for accessible methods to derive comprehensive spatial multiomics data from simpler single-omics measurements.

Purpose of the Study:

• To introduce NicheTrans, a novel spatially aware cross-omics translation method.
• To develop a flexible Transformer-based multimodal framework for enhanced spatial multiomics analysis.
• To demonstrate the utility of NicheTrans in uncovering previously undetectable spatial multiomics domains and facilitating broader multiomics adoption.

Main Methods:

• Developed NicheTrans, a Transformer-based multimodal framework incorporating cellular microenvironment and multimodal data.
• Validated NicheTrans across diverse biological datasets, comparing its performance to existing methods.
• Employed model interpretation techniques to identify key molecular relationships and spatial patterns.

Main Results:

• NicheTrans successfully uncovered spatial multiomics domains not identifiable through single-omics analysis alone.
• Identified key molecular relationships, including gene programs linked to dopamine metabolism and amyloid β-associated cell states.
• Quantified spatial organization of glial cell subtypes in Alzheimer's disease brain using translated protein markers.

Conclusions:

• NicheTrans effectively translates accessible single-omics data into comprehensive spatial multiomics insights.
• The method overcomes limitations of existing techniques by integrating microenvironment information.
• NicheTrans significantly enhances the feasibility and accessibility of multiomics analysis for the research community.