Toward Graph-Based Decoding of Tumor Evolution: Spatial Inference of Copy Number Variations

Area of Science:

• Genomics
• Computational Biology
• Cancer Research

Background:

• Tumor heterogeneity and evolution are critical areas in cancer research.
• Copy number variation (CNV) is a key feature of tumor heterogeneity.
• Existing methods for mapping CNV often ignore crucial spatial information.

Purpose of the Study:

• To develop a novel computational model for mapping tumor heterogeneity.
• To leverage spatial omics data for a comprehensive understanding of tumor evolution.
• To accurately infer copy number variations (CNVs) within the tumor microenvironment.

Main Methods:

• Introduction of SCOIGET (Spatial COpy number Inference by Graph on Evolution of Tumor).
• Utilizing graph neural networks with graph attention layers for spatial feature learning.
• Integration of spatial multi-omics data for enhanced tumor heterogeneity mapping.

Main Results:

• SCOIGET demonstrated superior performance with reduced error metrics and improved clustering compared to existing methods.
• The model accurately captures spatial and temporal changes in tumor evolution across various cancer types and spatial omics platforms.
• Validated on simulated data and patient cohorts, showing strong generalizability.

Conclusions:

• SCOIGET provides an innovative solution for detailed and accurate tumor heterogeneity and evolution mapping.
• The framework aids in understanding tumor progression and developing personalized cancer treatment strategies.
• Enhances research efficiency in cancer genomics and spatial biology.