NicheDeSig: Niche-aware Deconvolution and Adaptive Signature Analysis for Spatial Transcriptomics

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Accurate cell-type deconvolution is crucial for spatial transcriptomics (ST) spot analysis.
• Spatial niches provide context but are often unintegrated into deconvolution methods.
• Existing methods lack niche awareness, limiting functional analysis of spatial molecular programs.

Purpose of the Study:

• To develop a novel method, NicheDeSig, for niche-aware cell-type deconvolution in ST.
• To enable analysis of cell-state variations across distinct spatial micro-environments.
• To improve the functional interpretation of ST data by incorporating niche information.

Main Methods:

• NicheDeSig models cell types with adaptive signatures.
• It performs spot deconvolution using context-dependent signatures.
• The method incorporates niche priors and models niche-dependent signature shifts.

Main Results:

• NicheDeSig demonstrates strong deconvolution performance on benchmark datasets.
• The method enhances spatial fidelity in simulated and real ST data.
• Learned signatures reveal biological insights in human brain, breast cancer, and pancreatic cancer.

Conclusions:

• NicheDeSig advances ST analysis by integrating spatial niche information.
• The method supports more accurate cell-type deconvolution and functional interpretation.
• NicheDeSig provides a powerful tool for studying tissue micro-environments.