🧬 RxGenome

Explainable Disease Risk & Pharmacogenomic Insights

Built for the BACSA Hacks 2026 Open Challenge (Disease Detection & Health Improvement)

🎯 The Vision

At the intersection of technology and biology, RxGenome empowers patients and healthcare providers by turning raw, complex medical data into actionable biological insights.

Built in 24 hours for BACSA Hacks, RxGenome tackles the Open Challenge by addressing both Disease Detection and Health Improvement. It serves as a precision-medicine interpretation assistant—combining machine learning for genomic risk prediction, multimodal AI for report understanding, and evidence-based pharmacogenomic (PGx) rules to flag dangerous medication interactions.

⚠️ Educational Tool Only — RxGenome is a demonstration of applied health-tech and is not a diagnostic device. All results should be discussed with a qualified healthcare provider.

✨ Key Features

🧬 Explainable Disease Detection

• Machine Learning Analysis: Utilizes a trained XGBoost classifier over genomic/SNP-style feature matrices to accurately predict disease risk probability.
• Biological Context: Instead of black-box predictions, RxGenome provides an explainable Feature Importance chart, mapping statistical anomalies back to standard HGNC gene symbols (e.g., BRCA1, TP53, PIK3CA).

💊 Health Improvement via Pharmacogenomics (PGx)

• Medication Safety: Cross-references a patient's prescription data against FDA and CPIC guidelines.
• Metabolizer Alerts: Flags severe gene-drug interactions (e.g., CYP2D6 poor metabolizers taking Tamoxifen), generating specific discussion points to prevent adverse drug events.

📋 Multimodal Report Understanding

• AI-Powered Extraction: Integrates Google Gemini (2.0 Flash / 2.5 Pro) to physically read unstructured medical PDFs and lab report images.
• Smart Formatting: Automatically extracts vital lab values, identifies units and reference ranges, and flags abnormal biomarkers.

🩺 Clinician & Patient Summaries

• Patient-Friendly Breakdowns: Dynamically generates an empathetic, 8th-grade reading level summary of the patient's risk profile and PGx alerts.
• Doctor Discussion Notes: Auto-generates structured clinical notes, pulling together all extracted lab anomalies, model ROC-AUC confidences, and PGx cautions—equipping patients for their next appointment.

🛠️ The Tech Stack

RxGenome leverages a modern, decoupled architecture designed for speed, accuracy, and responsive UX.

🔬 How It Works

1. Input Data: The user uploads a unified case containing their genomic variant file (CSV/VCF), unstructured lab reports (PDF/Images), and current medication profiles.
2. Orchestration Pipeline: The FastAPI backend securely routes the data through four stages:
   • Genomic Parsing
   • Gemini OCR Document Extraction
   • Medication Normalization
   • XGBoost Risk Prediction
3. PGx Evaluation: The rule engine evaluates the identified genomic variants against the canonical medication sequence.
4. Synthesis: Gemini synthesizes the structured output of the ML model, the PGx rules, and the lab arrays into tailored conversational summaries.
5. Insights Dashboard: The React frontend renders a beautiful, interactive matrix of the patient's holistic health profile.

Built With

• fastapi
• gemini
• next.js
• pandas
• pydantic
• recharts
• scikit-learn
• sqlalchemy
• sqlite
• tailwindcss
• typescript
• xgboost