MultiOmics Research

Graph Attention
Autoencoders

Integrating high-dimensional oncology data modalities from TCGA—including genomics, transcriptomics, and epigenomics—into a unified latent space for nuanced cancer biomarker extraction and risk stratification.

Key Focus Areas

  • HNSCC Risk Prediction: Immunomethylomic Tuning of Pathology to Text Models to accurately forecast Head and Neck Squamous Cell Carcinoma.
  • Multiomic Integration: Translating raw TCGA sequencing data into spatial graphs for Attention Networks to capture structural biological connections.

HuggingFace Space

Interactive Demo

GitHub Repository

Source Code

Research Publication

Read the full methodology and metrics
Complete Project Flow
Omics Features
Node Features
Attention Result
Final Integration Result
Live View