Medical AI Experimentation Project

Overview

This repository contains information and resources for experimenting with medical AI, drug repurposing tools, and research automation based on a conversation about Every Cure's MATRIX platform and local medical AI possibilities.

Background

Inspired by David Fajgenbaum's work with Every Cure and the MATRIX platform, this project explores how to create local alternatives for medical research and drug repurposing analysis while we wait for the public release of professional-grade tools.

What's Included

• Summary - Key insights and actionable information from the conversation
• Resources for setting up local medical AI experiments

Getting Started with Local Medical AI

Prerequisites

• Computer capable of running large language models (8GB+ RAM recommended)
• Internet connection for downloading models and accessing research databases

Step 1: Install a Local LLM Application

Choose one of these user-friendly options:

• GPT4All (recommended for beginners)
• Jan.ai
• LM Studio

Step 2: Download Medical Models

Get GGUF format models from Hugging Face:

• BioMistral - General biomedical Q&A
• MedAlpaca - Clinical reasoning and patient cases
• Clinical LLaMA - Medical diagnosis conversations
• PMC-LLaMA - Biomedical research paper analysis

Step 3: Set Up Holistic Research Workflow

1. Define patient context: List all relevant factors (age, comorbidities, disabilities, lifestyle)
2. Search strategically: Use multiple PubMed searches:
   • "[drug] [primary condition]"
   • "[drug] [comorbidity] safety"
   • "[drug] [disability] interaction"
   • "[drug] smoking" (or other lifestyle factors)
   • OR automate with the PubMed API
3. Copy relevant abstracts from each search (2-3 per factor)
4. Use structured prompts that include all patient context factors

Example Workflow

Research Question Template

You are a biomedical research assistant.
Question: [Your specific question about drug repurposing/medical treatment]

Patient Context (if applicable):
- Age/demographics: [e.g., 45-year-old female]
- Comorbidities: [e.g., diabetes, hypertension, ADHD]
- Disabilities: [e.g., mobility limitations, cognitive disabilities]
- Lifestyle factors: [e.g., smoking, alcohol use, exercise level]
- Current medications: [list if relevant]
- Other considerations: [e.g., pregnancy, allergies, occupation]

Here are relevant PubMed abstracts:
[Paste 2-3 abstracts here]

Please provide:
- Summary of findings
- Possible biological mechanisms
- Known risks or unknowns
- Specific considerations for the patient context provided
- Drug interactions with current medications
- Lifestyle/disability modifications needed
- Contraindications based on comorbidities

Sample Questions to Explore

• "Could [existing drug] be repurposed for [specific condition] in patients with [comorbidity]?"
• "What are the safety considerations for [drug] in patients with [disability/condition] who also smoke?"
• "How does [treatment] effectiveness change in patients with multiple conditions like [condition A] and [condition B]?"
• "What treatment options exist for [rare disease] in patients with mobility disabilities?"
• "Are there drug interactions between [psychiatric medication] and [proposed treatment] for someone with ADHD and chronic pain?"

Useful Resources

Research Databases

• PubMed - Medical literature search
• Consensus.app - AI-powered research search engine
• Open Targets - Drug-disease associations and gene pathways

Professional Tools (For Reference)

• Every Cure MATRIX - Drug repurposing platform (public release planned 2026)
• BenevolentAI - Commercial drug discovery platform
• Healx - Rare disease drug repurposing

Safety and Legal Considerations

⚠️ Important Disclaimers:

• This is for research and educational purposes only
• Not medical advice - always consult healthcare professionals
• Local AI models are research assistants, not diagnostic tools
• Results should be verified through peer-reviewed literature

Future Possibilities

Short-term Goals

• Test different medical GGUF models
• Develop effective prompt templates for holistic patient analysis
• Create automated PubMed integration scripts
• Document successful research workflows
• Build templates for common comorbidity combinations
• Test multi-factor search strategies (drug + condition + disability)
• Create patient context checklists

Long-term Vision

• Build WikiLinks-style visualization for medical research
• Integrate multiple medical databases
• Create tools for holistic patient context analysis
• Contribute to open-source medical AI development

Contributing

If you experiment with these tools and discover effective workflows, please share your findings! This is a collaborative exploration of accessible medical AI.

Related Projects

• Every Cure - Professional drug repurposing platform
• Open Targets - Open-source drug-disease associations
• Bio-LLM projects - Various open-source medical language models

---

This project is inspired by the need for accessible medical research tools and the potential of AI to democratize drug discovery and repurposing research.