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

OTTER, an optimal-transport multi-omics deep learning framework, significantly outperforms state-of-the-art methods for T-lineage acute lymphoblastic leukemia subtyping in the COG AALL0434 cohort of >1,300 patients across 17 subtypes.

• Why it matters: T-ALL shows complex heterogeneity across transcriptomic and genomic layers, and existing diagnostics and computational tools are often slow or limited by single-modality or simplistic integration.
• What they did: OTTER uses modality-specific encoders for transcriptomic and genomic inputs and aligns their latent spaces with Gromov-Wasserstein optimal transport (GW-OT) to preserve each modality’s internal geometry without a shared feature space.
• The result: OTTER identifies interpretable subtype-defining and omics-specific features and generalizes across multiple external independent clinical cohorts, supporting its use for biologically interpretable multi-omics characterization.

AI summary of the abstract · may contain errors · not medical advice · read the paper.