Multi-omics
Multi-omics - Knowledge-Based CS: Mouse Human Candidate Gene Selection with Agentic RAG
Prioritize candidate genes by integrating mouse model data, literature, and human cohort validation.
Mentor Details:
Prof. Ilan Tsarfaty
Mentor Details:
Requirments:
Build explainable, evidence-driven gene ranking workflows.
Problem Statement
Translating findings from animal models to human disease remains challenging.
Project Objectives
Build explainable, evidence-driven gene ranking workflows.
Technical Scope
Create a practical pipeline for prioritizing candidate genes by integrating mouse variant signals, literature/knowledge retrieval, network biology, and validation in human cohorts (e.g., TCGA/METABRIC-style analyses). The emphasis is building an evidence-driven selection workflow that is both computationally rigorous and explainable.
Required Knowledge and Prerequisites
Core Requirements:
Python, data analysis.
Recommended Background:
Bioinformatics, graph analysis.
Project Difficulty and Expected Level
Overall Difficulty: varies
This project is well-suited for:
Teams of 2–4 students
Expected Outcomes
A rare “bridge skillset”: moving between AI, biological networks, and translational validation logic.
Experience with evidence-centric AI: building retrieval + synthesis flows that remain auditable and defensible.
A portfolio artifact that reads like applied R&D (not a classroom exercise): ranked gene lists, rationales, and validation plots.