RAIDers (Rare Disease & AI)

RAIDers is a federated computational framework for resolving the phenotypic heterogeneity of Amyotrophic Lateral Sclerosis (ALS) while preserving data sovereignty. By integrating global genomic annotations with mechanistic pathway analysis, this framework establishes a scalable architecture for rare disease molecular subtyping.

---

🎯 Scientific Objectives

• Subtype Discovery: Identify coherent molecular signatures across diverse ancestral backgrounds using federated machine learning
• Pathway-Level Analysis: Map genetic variants to biological mechanisms to understand disease heterogeneity
• Federated Feasibility: Demonstrate that analytical fidelity is maintained when data remains physically separated across institutional nodes
• Therapeutic Stratification: Enable precision medicine approaches by identifying distinct molecular subtypes requiring different therapeutic strategies

---

📊 Project Phases

Phase	Status	Description
Phase 1	✅ Complete	Federated K-means clustering for molecular subtype discovery
Phase 2	✅ Complete	Mechanistic pathway annotation and statistical analysis
Phase 3	🔄 In Progress	Multi-institutional collaboration and real-world validation

---

🧬 Phase 1: Federated Subtype Discovery

Synthetic Cohort Generation

To overcome the "mathematical invisibility" of rare variants, RAIDers employs a digital mutagenesis strategy generating a balanced cohort of 15,000 synthetic patients partitioned across five ancestral superpopulations (AFR, AMR, EAS, EUR, SAS; n=3,000 each).

Data Sources:

• ClinVar: ~450 pathogenic ALS variants across 34 genes
• gnomAD: Population-specific allele frequencies for ancestral modifiers

Federated K-Means Clustering

Molecular subtypes are discovered through a privacy-preserving decentralized algorithm:

1. Local Computation: Each population node computes cluster assignments using local patient data
2. Secure Aggregation: Only centroids (not patient data) are shared with a central coordinator
3. Global Update: Federated averaging produces updated global centroids
4. Convergence: Process repeats until centroid change < 0.001

Phase 1 Results

The algorithm identified 5 distinct clusters with clear clinical stratification:

Cluster	n	Mean Severity	Progression	Interpretation
C0	287	5.00	74% Slow	Mild
C1	3,287	7.27	93% Moderate	Moderate-A (SOD1-like)
C2	1,999	9.30	97% Fast	Severe
C3	8,957	0.00	N/A	Control (No variants)
C4	845	6.32	91% Moderate	Moderate-B (ALS2-like)

Key Finding: Clusters 1 and 4 both represent moderate severity but exhibit distinct molecular profiles, providing evidence for severity-stratified molecular subtypes.

---

🔬 Phase 2: Mechanistic Pathway Analysis

Pathway Annotation Framework

Phase 2 maps the 34 ALS-associated genes to 7 canonical molecular pathways:

Pathway	Genes	Representative
Proteostasis	8	SOD1, C9orf72, VCP, TBK1
RNA Metabolism	7	TARDBP, FUS, MATR3, HNRNPA1
Vesicle Trafficking	5	ALS2, CHMP2B, VAPB, FIG4
Cytoskeletal/Axonal	5	TUBA4A, PFN1, DCTN1, KIF5A
Mitochondrial	4	CHCHD10, SIGMAR1
DNA Damage	4	NEK1, SETX, SPG11
Excitotoxicity	4	UNC13A, DAO

Statistical Analysis Suite

Phase 2 implements comprehensive statistical methods:

• Co-occurrence Analysis: Odds ratios, Jaccard similarity, Fisher's exact test
• Correlation Analysis: Spearman ρ and Pearson r for pathway burden scores
• Quadrant Classification: Biological categorization of pathway relationships
• Cross-Cluster Comparison: Kruskal-Wallis tests with effect sizes (ε², Cohen's h)

Phase 2 Key Findings

1. Dose-Dependent Relationship

The Mitochondrial-Excitotoxicity axis emerged as the dominant mechanistic link:

• Co-occurrence: 86.5%
• Correlation: r = 0.687
• Odds Ratio: 38.78 (95% CI: 33.23–45.27)

This suggests cascading molecular failure where mitochondrial dysfunction drives excitotoxicity.

2. Distinct Molecular Subtypes

Proteostasis-Vesicle Trafficking mutual exclusivity:

• Co-occurrence: 24.4%
• Correlation: r = −0.408
• Odds Ratio: 0.11

Indicates these pathways represent separate disease aetiologies rather than co-occurring mechanisms.

3. Severity-Associated Gradients

Pathway	Trend (Mild → Severe)	Cohen's h
Proteostasis	↑ 0% → 73%	2.05 (Large)
RNA Metabolism	↑ 0% → 27%	1.09 (Large)
Vesicle Trafficking	↓ 46% → 14%	0.73 (Medium)
DNA Damage	↓ 42% → 10%	0.77 (Medium)

---

🚀 Quick Start

Installation

# Clone the repository
git clone https://github.com/your-org/RAIDers.git
cd RAIDers

# Install dependencies
pip install -r requirements.txt

Running the Full Pipeline

# Execute Phase 1 + Phase 2 pipeline
python main.py

Running Individual Components

# Phase 1: Federated Clustering
from fed_kmeans_components import FederatedKMeans

clusterer = FederatedKMeans(n_clusters=5, n_populations=5)
clusters = clusterer.fit(patient_features)

# Phase 2: Pathway Analysis
from phase2_pathway_annotation import PathwayAnnotator

annotator = PathwayAnnotator(gene_pathway_map='data/gene_pathway_map.json')
pathway_scores = annotator.annotate_patients(patient_variants)

Interactive Dashboard

Open als_pathway_dashboard_clusters.html in a web browser to explore:

• Pathway prevalence by cluster
• Co-occurrence heatmaps
• Correlation matrices
• Quadrant classification scatter plot
• Network visualization

---

📁 Repository Structure

RAIDers/
├── main.py                          # Main pipeline orchestrator
├── fed_kmeans_components.py         # Federated K-means implementation
├── phase2_pathway_annotation.py     # Pathway annotation and analysis
├── stage4_pathway_analysis.py       # Statistical analysis suite
├── generate_figures.py              # Publication figure generation
├── als_pathway_dashboard.html       # Interactive visualization
├── data/
│   ├── clinvar.cleaned.csv          # Curated pathogenic variants
│   ├── gene_pathway_map.json        # Gene-to-pathway mappings
│   └── outputs/                     # Analysis results (CSV, XLSX)
├── notebooks/
│   ├── total_pipeline.ipynb         # End-to-end tutorial
│   └── biological_validation.ipynb  # Validation analyses
└── docs/
    ├── METHODOLOGY.md               # Detailed methods
    └── figures/                      # Generated figures

---

🔮 Phase 3: Multi-Institutional Collaboration (In Progress)

Phase 3 extends RAIDers to real-world multi-institutional deployment:

Objectives

• Real Biobank Integration: Validation on controlled-access datasets (UK Biobank, All of Us)
• Swarm Learning: Transition from centralized to fully decentralized federated learning
• Clinical Validation: Correlation with longitudinal patient outcomes
• Therapeutic Targeting: Pathway-specific drug repurposing analysis

Collaboration Opportunities

We welcome collaborations with:

• Rare disease research consortia
• Biobank data custodians
• Pharmaceutical partners for therapeutic validation
• Computational biology groups

Contact: [See Contributors section below]

---

📈 Data Sources

Database	Purpose	Phase
ClinVar	Pathogenic variant curation	1, 2
gnomAD	Population allele frequencies	1
OMIM	Gene-disease associations	2
Orphanet	Rare disease ontologies	2, 3
Reactome	Pathway definitions	2

---

📚 Citation

If you use RAIDers in your research, please cite:

@software{raiders2026,
  title = {RAIDers: Federated Molecular Subtyping for Rare Disease},
  author = {Shah, Aastha and Kharbanda, Arnav and others},
  year = {2026},
  url = {https://github.com/your-org/RAIDers}
}

---

👥 Contributors

Name	Email	ORCID	Institution
Aastha Shah	aasthashah.work@gmail.com	0009-0008-7811-0177	Queen's University Belfast
Arnav Kharbanda	arnavkha@andrew.cmu.edu	0009-0007-9195-9960	Carnegie Mellon University
Bill Paseman	bill@rarekidneycancer.org	0000-0002-5020-0866
Chantera Lazard	lazard.c@northeastern.edu	0009-0006-1367-3812	Northeastern University
Jialan Ma	jialanma7@gmail.com	0009-0007-2670-9076	Broad Institute of MIT and Harvard
Kushal Koirala	kkoirala@unc.edu	0009-0009-7935-4533	University of North Carolina
Kyulin Kim	lynn.kim.24@ucl.ac.uk	0009-0007-8976-2405	University College London
Nikita Rajesh		0009-0009-9850-5261	Carnegie Mellon University
Pu (Paul) Kao	gaopuo1234@gmail.com	0009-0003-9047-0160	National Taiwan University
Shreya Nandakumar		0009-0006-9230-3659	Carnegie Mellon University
Vibha Acharya	via16@pitt.edu	0000-0001-6598-0052	University of Pittsburgh
William Lu	wtlu@andrew.cmu.edu	0000-0002-2768-1489	Carnegie Mellon University

---

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

---

Developed as part of the Rare Disease & AI Initiative