Multi-omics 7
Multi-omics: Agentic Multi-Task AI for Integrated Cancer Prognostics & Target Discovery
Integrate multi-omics and imaging data into unified predictive and discovery workflows using GenAI.
Mentor Details:
Prof. Ilan Tsarfaty
Mentor Details:
Requirments:
Develop multi-task AI models with evidence-grounded biomarker and target prioritization.
Problem Statement
Heterogeneous omics datasets are difficult to harmonize and interpret jointly.
Project Objectives
Develop multi-task AI models with evidence-grounded biomarker and target prioritization.
Technical Scope
Develop a multi-omics framework that integrates genomics, transcriptomics, proteomics, metabolomics, and imaging-derived features into unified predictive and discovery workflows. You’ll use GenAI + multi-task modeling and integrate Agentic RAG for evidence-grounded prioritization of biomarkers and therapeutic targets.
Required Knowledge and Prerequisites
Core Requirements:
Machine learning, data engineering.
Recommended Background:
Multi-omics analysis, causal ML.
Project Difficulty and Expected Level
Overall Difficulty: varies
This project is well-suited for:
Teams of 2–4 students
Expected Outcomes
A systems-level view of precision oncology: how heterogeneous omics are harmonized, fused, and validated.
Technical depth in modern applied ML: multi-task learning, causal-thinking patterns (where appropriate), and evidence-aware retrieval workflows.
Strong engineering practice: data versioning, modular services, and repeatable evaluation—skills that translate directly to industry ML teams.