AI-powered genetic variant analysis using cutting-edge genomic language models
GenomicsAI is a full-stack AI-powered web application that empowers researchers, clinicians, and curious learners to analyze the effects of genetic mutations on human health. Leveraging the state-of-the-art Evo 2 AI model, users can predict whether single nucleotide variants (SNVs) are benign or pathogenic with unprecedented accuracy.
- 🔍 Search for genes (e.g., BRCA1 for breast cancer research)
- 🧬 Explore genome sequences with interactive visualization
- ⚡ Input mutations (e.g., G → A at position X) for instant analysis
- 🤖 AI-powered predictions using the Evo 2 model for variant pathogenicity
- 📊 Compare predictions against real-world data from ClinVar database
- 📈 Confidence scoring to understand prediction reliability
This platform combines cutting-edge AI, biomedical research, and modern web technologies into a practical tool for genetic variant effect prediction.
| Layer | Technologies Used |
|---|---|
| 🧠 AI Model | Evo 2 - State-of-the-art genomic language model from Stanford et al. |
| ⚙️ Backend | Python, FastAPI, serverless hosted on Modal with H100 GPU |
| 🌐 Frontend | Next.js 15, React, TypeScript, Tailwind CSS, Shadcn UI |
| 📡 APIs | UCSC API (genome sequences), NCBI E-utilities, ClinVar |
Genetic mutations can significantly impact human health, potentially increasing risks for diseases like cancer, cardiovascular disorders, and neurological conditions. However:
- ❓ It's challenging to determine which mutations are harmful versus benign
- 🔬 Many variants are classified as "Variants of Unknown Significance" (VUS) with unclear clinical impact
- ⏰ Traditional validation methods require years of clinical studies
- 💰 Experimental characterization is expensive and time-consuming
💡 Our Solution: We harness Evo 2's deep learning capabilities, trained on 100,000+ genomes, to provide instant, accurate predictions of mutation pathogenicity.
Clinicians can evaluate whether patient-specific mutations are likely disease-causing, enabling faster diagnostic decisions and personalized treatment plans.
Scientists can rapidly assess novel mutations without waiting for extensive clinical validation, accelerating genomics research and drug discovery.
Genetic counselors can provide evidence-based risk assessments for inherited diseases, helping families make informed healthcare decisions.
Students and educators can explore complex genetic concepts through interactive analysis, making genomics more accessible and engaging.
Research Paper: "Genome modeling and design across all domains of life with Evo 2"
Institutions: Stanford University, UC Berkeley, UCSF, NVIDIA
Release Date: February 2025
Access: ResearchGate PDF
- 🌍 Massive Training Scale: Trained on 9.3 trillion nucleotides from over 128,000 species
- 🧬 Multi-functional: Predicts mutation impact, generates realistic DNA sequences, and enables genome design
- 🤖 Advanced Architecture: Transformer-based model similar to GPT, but optimized for biological sequences
- 🎯 State-of-the-art Accuracy: Achieves unprecedented precision in pathogenicity prediction
- 🔬 Research Impact: Enables AI-driven bioinformatics for simulation, prediction, and genome design
This groundbreaking research opens new frontiers in genomics, synthetic biology, and personalized medicine.
| Feature | Description |
|---|---|
| 🧬 Mutation Analysis | Input DNA mutations and receive AI-powered pathogenicity predictions |
| 🧠 AI vs. Clinical Data | Compare Evo 2 predictions with validated ClinVar database entries |
| 📊 Confidence Estimation | Understand prediction reliability with confidence scores |
| 🌍 Genome Assembly Support | Choose between genome versions (hg19, hg38, etc.) |
| 🔍 Advanced Gene Search | Search by gene name or browse by chromosomal location |
| 📋 Reference Sequence Viewer | Interactive display of complete gene sequences via UCSC API |
| 🧪 Variant Database Integration | Browse known mutations and their clinical significance |
| ⚡ High-Performance Backend | Real-time GPU-accelerated analysis via FastAPI and Modal |
| 🎨 Modern UI/UX | Responsive, intuitive interface built with Tailwind CSS and Shadcn UI |
- Node.js 18+
- Python 3.9+
- Modal account (for GPU backend)
# Clone the repository
git clone https://github.com/sehajmakkar/GenomicsAI.git
cd genomicsai
# Install frontend dependencies
npm install
# Install backend dependencies
pip install -r requirements.txt
# Set up environment variables
cp .env.example .env
# Add your API keys and configuration
NEXT_PUBLIC_ANALYZE_SINGLE_VARIANT_BASE_URL = ""
# Start the development servers
npm run dev # Frontend (Next.js)
modal run main.py # Backend (Modal + FastAPI)We welcome contributions! Please see our Contributing Guidelines for details on:
- Code standards and formatting
- Testing requirements
- Pull request process
- Issue reporting
This project is licensed under the MIT License - see the LICENSE file for details.
- Stanford AI Lab for the Evo 2 model and research
- NCBI for ClinVar and genomic databases
- UCSC Genome Browser for sequence data APIs
- Modal Labs for GPU infrastructure
- Vercel for frontend hosting
- Project Lead: Sehaj Preet
- Issues: GitHub Issues
- Discussions: GitHub Discussions
⭐ Star this repository if GenomicsAI helps advance your genomics research!
Built with ❤️ for the scientific community