Biology-Aware RAG for Target Discovery

AI-powered spatial transcriptomics analysis platform for therapeutic target validation at Genentech Research & Early Development.

Overview

This application demonstrates a Retrieval-Augmented Generation (RAG) tool specifically designed for In Vivo workflow insights. It allows scientists to query six months of spatial transcriptomics data to identify outliers in therapeutic target validation.

Key Features

• Natural Language Queries: Ask questions about spatial transcriptomics data in plain English
• Biology-Aware Context Retrieval: Intelligent retrieval of relevant experimental data
• Outlier Detection: Automatic identification of unexpected findings in target validation
• Multi-Model Support: Data from Alzheimer's (APP/PS1), HER2+ Breast Cancer, Pancreatic Cancer (KRAS), and other therapeutic areas

Getting Started

Prerequisites

• Node.js 18+
• OpenAI API key

Installation

npm install

Configuration

1. Copy .env.example to .env.local:

   cp .env.example .env.local

2. Add your OpenAI API key to .env.local:

   OPENAI_API_KEY=your_key_here
   

Running the Application

npm run dev

The app will be available at http://localhost:3002

Running Tests

npm run test:e2e

Example Queries

• "What are the outliers in therapeutic target validation?"
• "Show me findings from the APP/PS1 Alzheimer's model"
• "What spatial patterns exist in HER2+ breast cancer data?"
• "Summarize KRAS findings in the pancreatic cancer model"
• "Which targets have the highest validation scores?"

Data Overview

The demo includes mock spatial transcriptomics data representing:

• 14 experimental records across 4 mouse models
• 6 therapeutic targets with validation scores
• 5 outlier findings with actionable insights

Mouse Models

Model	Indication	Key Targets
C57BL/6-APP/PS1	Alzheimer's Disease	TREM2, APOE, BACE1, MAPT
BALB/c-HER2	HER2+ Breast Cancer	ERBB2, CDK4, PIK3CA, PD-L1
Pdx1-Cre-LSL-KrasG12D	Pancreatic Cancer	KRAS, TP53, MET
NOD-SCID-IL2Rg	Immuno-Oncology	CD19, BCMA

Technology Stack

• Framework: Next.js 16
• Language: TypeScript
• AI: OpenAI GPT-4
• Testing: Playwright

Strategic Value

This application demonstrates Quome's capability to build "biology-aware" systems that:

1. Bridge lab execution and high-performance compute
2. Deliver real-world ROI through actionable insights
3. Support "Lab-in-the-Loop" workflows
4. Ensure FAIR data principles in research

---

Built with Quome