arrow_back HD Research Hub

What Can an LLM Learn from 22 HD Research Papers?

Experiment #1

Experiment Card

Experiment ID
EXP-001
Date / Time
2026-03-27T23:25:23Z (UTC-5)
Type
Exploratory / Hypothesis Generation
Status
Complete
Hypothesis
An open-source LLM can identify cross-paper patterns and generate drug repurposing starting points from HD research abstracts
Outcome
Partially confirmed. Cross-paper patterns found (TDP-43, PTMs). But model lacks clinical trial history knowledge.
Infrastructure
Model: Llama 3.1 8B
Runtime: Ollama 0.16.1
Hardware: NVIDIA Jetson AGX Orin 64GB
GPU: Ampere, 2048 CUDA cores
OS: Ubuntu 22.04 / JetPack 6.2.2
Duration: ~25 minutes total
Cost: $0 (local inference)
Cloud: None
Data
Source: PubMed E-utilities API
Queries: 5 search terms
Window: 90 days (2026-01 to 2026-03)
Input: 22 paper abstracts
Parsed: 20/22 (2 failed)
Output: JSON + Markdown report
Data license: Public domain (NIH)
Code license: MIT
Reproducibility
git clone https://github.com/jravinder/hd-research-agent && cd hd-research-agent && pip install -r requirements.txt && python src/run_experiment.py
The Question

If we feed recent HD research papers to an open-source LLM running on a $1,000 edge device, what patterns does it find?

What We Did

library_books

1. Pull Papers

22 papers from PubMed across 5 queries: treatment, AI/ML, somatic expansion, repurposing, biomarkers.

memory

2. Run LLM

Llama 3.1 8B on NVIDIA Jetson AGX Orin 64GB via Ollama. Local inference, zero cloud cost.

biotech

3. Extract

Targets, compounds, key findings, and repurposing signals from each abstract.

lightbulb

4. Hypothesize

Which FDA-approved drugs might work for HD?

Results

20
Analyzed
17
High Relevance
7
Novel Targets
5
Hypotheses

Top Hypotheses

IL-6 Pathway

Tocilizumab

Confidence80/100

IL-6 receptor blocker (arthritis drug). May reduce neuroinflammation from mutant huntingtin.

Glutamate

Riluzole

Confidence60/100

Already tested in HD with modest results. Model didn't know this. AI needs human verification.

TDP-43 / GSK-3β

Lithium

Confidence55/100

Novel TDP-43/HD connection. Cross-paper insight that's hard for humans to spot quickly.

What Worked

  • check_circle

    Cross-paper patterns

    Found TDP-43/HD connection and post-translational modification targets.

  • check_circle

    Categorization

    Correctly categorized 20/22 papers by research type.

  • check_circle

    Zero cost

    Ran on a Jetson in a home office. Repeatable nightly.

What Didn't

  • cancel

    No trial history

    Suggested Riluzole without knowing it was already tested.

  • cancel

    Biomarker confusion

    Confused NfL (biomarker) with a causal target.

  • cancel

    2 parse failures

    Abstracts too short for the model.

warning

Limitations

  • • Not reviewed by HD domain experts or clinicians
  • • No experimental validation
  • • Abstracts only, not full texts
  • • Single model (Llama 3.1 8B), single run
  • • Scores reflect model self-assessment, not clinical probability

Methodology

01

Data

PubMed E-utilities API, 5 queries, 90-day window

02

Model

Llama 3.1 8B, Ollama, NVIDIA Jetson AGX Orin 64GB

03

Analysis

Structured JSON extraction per paper, cross-paper synthesis

View commit history on GitHub

Back to HD Research Hub