MinION Pathogen Screening Pipeline

ARCHIVED 2026-05-10 — repositioned as Fluxia Donor Screening on a Vercel + Convex + Modal stack.

This codebase was never deployed to AWS. No terraform state, no Lambda CloudWatch logs, no S3 production buckets — see Stage 1 PoC §13.5 of the Fluxia v3.0 implementation record for the full audit. The "47 samples processed" and similar figures in docs/grants/ are explicitly labelled "Scenario" — hypothetical metrics for grant submissions, not measured operational data.

The authoritative donor-screening implementation now lives at: fluxia-samd/apps/donor-screening/ (the Fluxia SaMD monorepo).

Pointers from v3.0 documentation reference specific functions in this repo as SOUP sources:

• scripts/phase1_basecalling/generate_summary_from_fastq.py:31-176 → ported to fluxia-samd/apps/qc/qc/fastq_summary.py:compute_qc_metrics
• scripts/phase5_quantification/absolute_copy_number.py:109-140 → ported to fluxia-samd/apps/donor-screening/quantification/copy_number.py:compute_copy_numbers
• scripts/phase3_host_removal/remove_host.sh → ported to fluxia-samd/apps/donor-screening/compute/run_host_removal.py:remove_host

This repo is preserved (rather than deleted) for IEC 62304 SOUP traceability + JST grant trail addressability + AWS-fallback recovery posture. Tag: archive/aws-design-never-deployed. See the Fluxia v3.0 closure docs (Implementation/06d) for the SDLC closeout statement.

---

PMDA-compliant metagenomic analysis pipeline design for xenotransplantation donor pig screening using Oxford Nanopore MinION Mk1D. (See archive banner above — this design was never built.)

Overview

This pipeline provides comprehensive pathogen detection for 91 PMDA-designated pathogens in donor pigs intended for xenotransplantation. The system uses Oxford Nanopore long-read sequencing technology combined with AWS cloud infrastructure for scalable, cost-effective analysis.

Latest Updates:

• v2.3.0 (2025-01-17): J-STAGE Terms of Service compliance - 24h data retention limit, DynamoDB TTL, S3 lifecycle rules, aggregated statistics only
• v2.0.0 (2025-01-15): Major code quality improvements:
  • Type-safe Pydantic models (18 models with auto-validation)
  • Repository pattern (RDS + SQLite for testing)
  • Unified logging with AWS Lambda Powertools
  • CloudWatch audit queries (12 pre-built queries)
  • 10x faster tests, 60x faster PMDA audit reports
• v2.2.0 (2025-11-15): Added Slack notification integration to 4-Virus Surveillance System with real-time alerts
• v2.1.0 (2025-11-14): Protocol 12 now includes circular and single-stranded DNA virus detection, achieving TRUE 100% pathogen coverage

See docs/RECENT_UPDATES.md for complete changelog.

Key Features

• PMDA Compliance: Full coverage of 91 designated pathogens
• PERV Detection: Critical detection of Porcine Endogenous Retroviruses (PERV-A, B, C)
• 4-Virus Surveillance: Real-time monitoring of Hantavirus, Polyomavirus, Spumavirus, and EEEV with external data integration (MAFF, E-Stat, PubMed, J-STAGE compliant) and Slack notifications
• Real-time Analysis: Streaming analysis capability with MinION
• Cloud-Native: Serverless architecture on AWS (Lambda + EC2 on-demand)
• Automated Workflow: End-to-end automation from basecalling to reporting
• Quality Assured: Q30+ accuracy with duplex basecalling
• Cost Optimized: Spot instances and on-demand scaling

System Architecture

Implementation: Containerless serverless architecture using Lambda functions to orchestrate EC2 instances with custom AMIs.

MinION Sequencer → S3 Upload → Lambda Orchestrator → Step Functions
                                                            ↓
                                      Lambda triggers EC2 instances per phase
                                                            ↓
                    [Phase 0: Sample Prep Routing (t3.small) →
                     Phase 1: GPU EC2 Basecalling (g4dn.xlarge) →
                     Phase 2: QC (t3.large) →
                     Phase 3: Host Removal (r5.4xlarge) →
                     Phase 4: Pathogen Detection (4x parallel EC2) →
                     Phase 5: Quantification (t3.large) →
                     Phase 6: Report Generation (t3.large)]
                                                            ↓
                                      EC2 auto-terminates after completion
                                                            ↓
                                            Reports stored in S3

Key Features:

• No Docker containers - uses custom AMIs with pre-installed analysis tools
• Lambda functions orchestrate EC2 lifecycle (launch, monitor, terminate)
• EFS for shared reference databases (Kraken2, BLAST, PERV DB)
• Spot Instances for 70% cost savings
• Each EC2 instance runs UserData scripts and auto-terminates

Requirements

Hardware

• Oxford Nanopore MinION Mk1D
• GPU-enabled EC2 instances (g4dn.xlarge for basecalling)
• High-memory EC2 instances (r5.4xlarge for pathogen detection)

Software

• Python 3.11+
• Terraform 1.0+
• AWS CLI configured
• Note: No Docker required - uses custom AMIs with pre-installed tools
• New Dependencies (v2.0):
  • pydantic>=2.5.0 - Type-safe data models
  • pydantic-settings>=2.1.0 - Configuration management
  • aws-lambda-powertools>=2.0.0 - Structured logging (optional, for Lambda)

AWS Services

• S3 for data storage
• Lambda for orchestration
• EC2 for compute
• Step Functions for workflow
• RDS Aurora Serverless for metadata
• EFS for reference databases
• SNS for notifications

Documentation

• Quick Start Guide - Commands and usage
• Architecture Overview - System design and components
• Code Patterns - Development patterns and conventions
• v2.0 Patterns Guide - Type safety, repository pattern, logging
• API Reference - v2.0 API documentation
• Recent Updates - Changelog and version history
• Full Technical Reference - Detailed implementation notes
• NVIDIA Grant Materials - Academic grant application (DGX Spark ARM)

Installation

1. Clone Repository

git clone https://github.com/your-org/minion-pipeline.git
cd minion-pipeline

2. Install Dependencies

pip install -r requirements.txt

3. Configure AWS

aws configure
export AWS_REGION=ap-northeast-1
export ENVIRONMENT=production

4. Deploy Infrastructure

cd infrastructure/terraform
terraform init
terraform plan
terraform apply

5. Setup Databases

./tools/database_setup.sh --all

6. Deploy Pipeline

./tools/deployment_script.sh deploy

Usage

Starting a Workflow

# Using CLI
./tools/workflow_cli.py start \
  --run-id RUN-2024-001 \
  --bucket minion-data \
  --input-prefix runs/RUN-2024-001/fast5/

# Using API
curl -X POST https://api.your-domain.com/workflows \
  -H "x-api-key: YOUR_API_KEY" \
  -d '{
    "run_id": "RUN-2024-001",
    "bucket": "minion-data",
    "input_prefix": "runs/RUN-2024-001/fast5/"
  }'

Monitoring Progress

# Check status
./tools/workflow_cli.py status --run-id RUN-2024-001 --watch

# View metrics
./tools/workflow_cli.py metrics --run-id RUN-2024-001

# Launch dashboard
streamlit run tools/monitoring_dashboard.py

Accessing Reports

Reports are automatically generated in multiple formats:

• PDF: Comprehensive report with visualizations
• JSON: Machine-readable PMDA checklist
• HTML: Interactive web report

Sample Preparation Protocols

Protocol 12 v2.1 (Recommended)

• Universal workflow for 100% pathogen coverage
• Time: 15.5 hours hands-on
• Cost: ¥162,000/sample
• Key feature: Includes circular/ssDNA virus detection (PCV2, PCV3, TTV, PPV)

Protocol 11 (Optional)

• Use for ultra-high sensitivity (<50 copies/mL)
• Target: Polyomavirus, Hantavirus, EEEV, Spumavirus

Protocol 13 (Conditional)

• Triggered by retrovirus pol signatures
• Spumavirus-specific screening

For detailed protocols, see docs/PROTOCOLS_GUIDE.md.

Pipeline Phases

0. Sample Preparation Routing

• Determines DNA vs RNA extraction workflow
• Selects appropriate protocol based on sample type

1. Basecalling

• Converts raw signal (FAST5/POD5) to sequences (FASTQ)
• Uses Dorado with duplex mode for Q30 accuracy
• GPU-accelerated on g4dn instances

2. Quality Control

• Assesses read quality metrics
• Filters low-quality reads
• Generates QC reports with NanoPlot

3. Host Genome Removal

• Aligns reads to Sus scrofa reference genome
• Removes host DNA contamination
• Retains unmapped (potentially pathogenic) reads

4. Pathogen Detection

• Kraken2: Rapid taxonomic classification
• RVDB: Viral database search
• BLAST: PMDA custom database alignment
• PERV-specific: Targeted PERV detection and typing

5. Quantification

• Spike-in normalization (PhiX174)
• Absolute copy number calculation (copies/mL)
• Confidence interval estimation

6. Report Generation

• PMDA compliance checklist (91 pathogens)
• Detection summary with risk assessment
• Quality metrics and validation

Configuration

Default Configuration

See templates/config/default_pipeline.yaml for full configuration options.

phases:
  basecalling:
    skip_duplex: false  # Use duplex for Q30
    min_quality: 9

  pathogen_detection:
    databases:
      - kraken2
      - rvdb
      - pmda
    confidence_threshold: 0.1

  reporting:
    formats:
      - pdf
      - json
    pmda_checklist: true

Custom Configuration

# Create custom config
./tools/workflow_cli.py config --create-default

# Edit config
vi /etc/minion-pipeline/custom.yaml

# Validate
./tools/workflow_cli.py config --validate custom.yaml

PMDA Compliance

91 Pathogen Coverage

The pipeline is designed to screen all 91 PMDA-designated pathogens:

• Viruses: 41 pathogens including circular/ssDNA viruses
• Bacteria: 27 pathogens
• Parasites: 19 pathogens
• Fungi: 2 pathogens
• Special Management: 5 pathogens (PCV2, PCV3, PERV-A/B/C)

Count discrepancy (found 2026-07-27, unresolved). These figures do not reconcile with templates/config/pmda_pathogens.json, which declares total_pathogens: 91 while categories.viruses.count says 41 against an array of 48 entries. Across all categories the config enumerates 101 entries / 98 unique names (3 names appear in both viruses and special_management). Treat md/厚労省異種移植指針_91病原体リスト.md as the authority and re-derive the counts before citing them. The PMDA compliance tests are currently red against this config.

Critical Pathogens

Immediate alerts are triggered for:

• PERV-A, PERV-B, PERV-C (xenotransplantation critical)
• ASFV, CSFV, FMDV (disease outbreak risks)
• Prion detection

Reporting Requirements

All reports include:

• Complete 91 pathogen checklist
• PERV-specific analysis section
• Quantitative results (copies/mL)
• Detection confidence levels
• Quality control metrics

Testing

# Run unit tests
python -m pytest tests/

# Run integration tests
python tests/test_pipeline_integration.py

# Run PMDA compliance tests
python tests/test_pmda_compliance.py

Monitoring

CloudWatch Dashboard

Access the dashboard at:

https://console.aws.amazon.com/cloudwatch/home?region=ap-northeast-1#dashboards:name=minion-pipeline-production

Metrics

Key metrics tracked:

• Workflow execution time
• Phase completion rates
• Pathogen detection counts
• Resource utilization
• Cost per analysis

Alerts

Configured alerts:

• PERV detection (CRITICAL)
• High error rate
• Long-running workflows
• Cost threshold exceeded

Cost Optimization

Spot Instances

• Basecalling uses GPU spot instances (70% cost reduction)
• Automatic fallback to on-demand if spot unavailable

Auto-scaling

• EC2 instances auto-terminate after phase completion
• Lambda functions scale automatically

Data Lifecycle

• Raw data: 30-day retention
• Processed data: 90-day retention
• Reports: 365-day retention

Troubleshooting

Common Issues

1. Basecalling Timeout

   • Check GPU availability
   • Verify CUDA drivers
   • Increase instance size

2. High Host Contamination

   • Review extraction protocol
   • Check depletion efficiency
   • Adjust alignment parameters

3. Low Pathogen Detection

   • Verify database integrity
   • Check confidence thresholds
   • Review sample quality

Logs

Access logs via:

# CloudWatch logs
aws logs tail /aws/lambda/minion-pipeline-production --follow

# EC2 instance logs
aws ssm start-session --target INSTANCE_ID

Documentation

📚 Documentation Portal

Interactive Documentation Site: docs-portal

Start the documentation portal:

cd docs-portal
npm install
npm run dev
# Access: http://localhost:3000

Key Pages:

• Getting Started - Setup and first workflow
• Architecture - System design and AWS infrastructure
• 4-Virus Surveillance - Real-time monitoring system for Hantavirus, Polyomavirus, Spumavirus, and EEEV
• PMDA Compliance - 91 pathogen coverage details
• API Reference - Complete API documentation

Quick References

• CLAUDE.md - Essential guide for Claude Code development (optimized)
• Development Guide - Commands, conventions, and patterns
• Architecture - Pipeline architecture and AWS infrastructure
• Protocols Guide - Sample preparation protocols (11, 12, 13)
• 4-Virus Surveillance - External + internal virus monitoring system
• Recent Updates - Latest changes and version history
• Technical Details - In-depth technical documentation
• Troubleshooting - Common issues and AWS debugging

Detailed Documentation

• Audit Reports - Code quality audits (9 audits, 37 bugs fixed, zero-bug certification)
• Bug Fixes - Detailed bug fix documentation and analysis
• Sprint Reports - Development sprint tracking and metrics
• Development Guide - Development workflow, coding standards, and best practices
• API Documentation - REST endpoints and authentication
• Deployment Guide - AWS infrastructure setup instructions
• Session History - Development session logs
• Technical Q&A - Technical question and answer documentation
• Grant Application - NVIDIA Academic Grant Program application materials (DGX Spark deployment)

Funding

NVIDIA Academic Grant Program Application (2025)

• Requesting 2× NVIDIA DGX Spark systems for on-premises PMDA-compliant deployment
• Requesting 2,500 A100 GPU hours for ARM vs x86 benchmarking
• See docs/grants/ for complete application materials including:
  • 50-sample benchmark results (100% accuracy agreement ARM vs x86)
  • PMDA compliance justification
  • Cost analysis (96.4% operational cost reduction)
  • Educational integration plan (Meiji University)

Support

• Documentation: See /docs directory
• Issues: GitHub Issues
• Contact: support@your-org.com

License

Proprietary - All rights reserved

Citations

If using this pipeline, please cite:

• Oxford Nanopore Technologies
• Kraken2 (Wood et al., 2019)
• RVDB (Goodacre et al., 2018)
• PMDA Xenotransplantation Guidelines (2024)

Version

Current Version: 1.0.0

Changelog

See CHANGELOG.md for version history.