Molecular Characterization of T-Lineage Acute Lymphoblastic Leukemia by an Optimal-Transport Based Multi-Omics...

Area of Science:

• Computational biology
• Genomics
• Pediatric oncology

Background:

• T-lineage acute lymphoblastic leukemia (T-ALL) is a complex pediatric cancer with significant molecular heterogeneity.
• Current subtyping methods are time-consuming and limited by single-data sources.
• Integrating multi-omics data for T-ALL characterization presents a computational challenge.

Purpose of the Study:

• To develop a novel computational framework for integrating multi-omics data in T-ALL.
• To improve molecular characterization and subtyping of T-ALL.
• To identify subtype-driving molecular features and cross-modal interactions.

Main Methods:

• Developed OTTER (Optimal Transport-based Transcriptomics and gEnomics Representation fusion), a multi-modal deep learning framework.
• Utilized variational autoencoders for modality-specific encoding and Gromov-Wasserstein optimal transport (GW-OT) for latent representation alignment.
• Applied the framework to the COG AALL0434 cohort (1,309 patients) and performed gradient-based interpretability analyses.

Main Results:

• OTTER effectively integrates RNA-seq and genomic variant data for T-ALL molecular characterization.
• Identified subtype-driving molecular features and coordinated cross-omics programs.
• Demonstrated the framework's ability to uncover the coordinated molecular landscape of T-ALL.

Conclusions:

• OTTER offers a principled, interpretable, and effective computational framework for multi-omics T-ALL characterization.
• The GW-OT approach enables geometry-preserving cross-modal alignment.
• The framework is generalizable to other cancers and multi-omics integration tasks.