Deep Subspace Reconstruction-Driven Adaptive Canonical Correlation Analysis Model for Identifying Imaging Genetics...

Area of Science:

• Computational Biology
• Bioinformatics
• Medical Imaging Analysis

Background:

• Clinical data often overlooks crucial information in raw data manipulation.
• Canonical correlation analysis extensions require advanced models for multi-modal data integration.
• Understanding sarcoma recurrence necessitates integrating diverse data types like imaging and genetics.

Purpose of the Study:

• To develop a novel two-stage correlated analysis model for integrating whole slide images, gene expression, and pathway scoring data.
• To identify high-order correlations linked to sarcoma recurrence by overcoming limitations in existing methods.
• To uncover potential biological mechanisms and pathways associated with sarcoma recurrence for personalized treatment strategies.

Main Methods:

• Utilized a deep self-reconstruction (DSR) model for multi-modal data integration.
• Employed hypergraph-based adaptive sparse multi-view canonical correlation analysis (AdaSMCCA) to explore higher-order correlations.
• Integrated data from 259 sarcoma samples from TCGA and UCSC databases.

Main Results:

• The DSR-AdaSMCCA model demonstrated faster convergence and higher correlation coefficients compared to existing methods.
• Bioinformatics analysis identified genes enriched in sarcoma recurrence-related diseases.
• The study successfully linked whole slide image features with genetic characteristics, revealing pathways in immune response and inflammation.

Conclusions:

• The proposed model effectively integrates imaging genetics data to identify key features for sarcoma local recurrence.
• The findings provide valuable insights into the complex networks influencing tumor relapse and personalized treatment strategies.
• The study deepens the understanding of sarcoma recurrence by uncovering potential biological mechanisms and therapeutic targets.