Genome Biology 2024 Explainable Multi-omics Pretraining

TMO-Net

An explainable pretrained multi-omics model for transferable oncology prediction and pathway-level interpretation.

TMO-Net showcase image

Overview

TMO-Net is designed for oncology settings where researchers need multi-omics models that are both transferable and interpretable. The project focuses on pretraining reusable biological representations across tasks while preserving the ability to trace predictions back to meaningful pathways and mechanisms.

Instead of treating multi-omics prediction as a pure black box, TMO-Net positions explainability as part of the model design. It extends the lab’s omics direction from robust classification toward reusable and mechanistically informative representation learning.