Skip to content

README.md

Latest commit

 

History

History
525 lines (405 loc) · 22.7 KB

README.md

File metadata and controls

525 lines (405 loc) · 22.7 KB
Raw

PAT002-BC-2026: Synthetic Patient Data

Overview

This directory contains 100% synthetic patient data created for demonstration and testing purposes. The data represents a realistic but completely fabricated Stage IIA ER+/PR+/HER2- Invasive Ductal Carcinoma (IDC) breast cancer patient case with BRCA2 germline mutation.

Patient: Michelle Anne Thompson (synthetic) Diagnosis: Stage IIA (T2N0M0) ER+/PR+/HER2- Breast Cancer, BRCA2+ (c.5946delT) Treatment Status: Post-adjuvant therapy (surgery + chemotherapy + radiation), currently on tamoxifen Disease Status: Disease-free at surveillance (2026-01-15) Purpose: Demonstrate precision medicine MCP server workflows for breast cancer

Data Origin & Creation

⚠️ CRITICAL: ALL DATA IS SYNTHETIC

This is NOT real patient data. All files were artificially generated using:

  1. Statistical models - Random number generators with clinically plausible distributions
  2. Expert knowledge - Curated to reflect realistic breast cancer biology (Luminal A/B subtype)
  3. Published literature - Based on TCGA-BRCA studies, BRCA2 mutation data, and clinical guidelines
  4. No PHI/PII - Contains zero protected health information or personally identifiable information

Source: Created by the Precision Medicine MCP development team for demonstration purposes (2026-01-31).

Data Modalities (5 types, 39+ files)

1. Clinical Data (clinical/)

File Size Description Source
patient_demographics.json 6.5 KB Patient demographics, BRCA2 status, family history Synthetic
lab_results.json 7.2 KB CEA/CA 15-3 tumor markers, CBC, liver/lipid panels Modeled after clinical patterns
fhir_raw/patient-002.json 2.1 KB FHIR R4 Patient resource Synthetic FHIR-compliant
fhir_raw/condition-breast-cancer.json 2.3 KB FHIR Condition resource (diagnosis) Synthetic
fhir_raw/observation-cea.json 1.8 KB CEA tumor marker observations Synthetic
fhir_raw/observation-brca2.json 2.0 KB BRCA2 genetic test result Synthetic
fhir_raw/observation-er-pr-her2.json 2.5 KB Hormone receptor status (ER 85%, PR 70%, HER2 neg) Synthetic
fhir_raw/medication-tamoxifen.json 1.7 KB Active tamoxifen prescription Synthetic
fhir_raw/medication-doxorubicin.json 1.6 KB Completed AC chemotherapy Synthetic
fhir_raw/medication-paclitaxel.json 1.6 KB Completed taxane chemotherapy Synthetic

Generation method:

  • Demographics: Randomly generated name (Michelle Thompson, 42F), dates, identifiers
  • Family history: Maternal breast cancer (age 58), maternal aunt ovarian cancer (age 52), sister BRCA2+ carrier
  • CEA/CA 15-3 values: Gaussian noise around normal post-treatment surveillance trajectory
  • FHIR resources: Standards-compliant with proper SNOMED CT, ICD-10, LOINC, RxNorm codes

Clinical scenario:

  • Diagnosis: December 2024 (Stage IIA, T2N0M0, 2.5cm tumor, node-negative)
  • Surgery: January 2025 (lumpectomy + sentinel node biopsy)
  • Chemotherapy: February-May 2025 (AC-T: doxorubicin/cyclophosphamide → paclitaxel)
  • Radiation: June-July 2025 (adjuvant)
  • Current: Tamoxifen 20mg daily (started August 2025, planned 5-10 years)
  • Surveillance: Q3 months with tumor markers, imaging annually

2. Genomic Data (genomics/)

File Size Description Source
somatic_variants.vcf 2.8 KB Germline BRCA2 + somatic mutations (original) Based on TCGA-BRCA + ClinVar
PAT002_somatic.vcf 1.6 KB Patient-prefixed somatic variants (PIK3CA/GATA3/CDH1/MAP3K1/TP53) Synthetic
copy_number_results.cns 0.7 KB CNVkit segmentation (original) Synthetic
PAT002_cnv.cns 1.3 KB Patient-prefixed BC CNV profile (25 segments, all autosomes) Synthetic

Generation method:

  • Selected common breast cancer driver mutations from TCGA-BRCA dataset
  • VCF format (v4.2) with realistic variant quality scores
  • Germline BRCA2 c.5946delT frameshift mutation (heterozygous in tumor and normal)
  • Somatic mutations in tumor only
  • Patient-prefixed files generated by generate_PAT002_missing_files.py

Key mutations (realistic for BRCA2+ ER+ breast cancer):

Germline:

  1. BRCA2 c.5946delT (chr13:32339811, frameshift) - GT: 0/1, AF: 0.50 in both tumor/normal
    • COSMIC: COSV57492880 (pathogenic, Class 5)
    • Homologous recombination deficiency (HRD) phenotype
    • PARP inhibitor eligible

Somatic (tumor-only): 2. PIK3CA H1047R (chr3:178952085, G>A) - AF: 0.42

  • COSMIC: COSM775 (hotspot mutation)
  • Common in ER+ breast cancer (~40% prevalence)
  • PI3K/AKT/mTOR pathway activation

  1. GATA3 frameshift (chr10:8100656) - AF: 0.38

    • Tumor suppressor in breast epithelium

  2. MAP3K1 missense (chr5:56111569) - AF: 0.35

    • Loss-of-function, favorable prognosis in ER+ BC

  3. MYC amplification (chr8:128748315) - Copy number 6

    • Proliferation driver

  4. CCND1 amplification (chr11:69455873)

    • Common in ER+ breast cancer

  5. TP53 wild-type - No mutation (typical for Luminal A/B, not TNBC)

  6. ESR1 wild-type - No mutation (tamoxifen-sensitive, no resistance)

  7. ERBB2 (HER2) wild-type - No amplification (HER2-negative)

Patient-prefixed somatic VCF (PAT002_somatic.vcf):

  • PIK3CA p.H1047R (chr3:179234297, A→G, VAF 0.42) — most common ER+ BC driver
  • GATA3 frameshift (chr10:8095656, VAF 0.31) — common in luminal BC
  • CDH1 splice donor (chr16:68771967, G→A, VAF 0.28) — common in lobular BC
  • MAP3K1 p.Q761X nonsense (chr5:56798505, C→T, VAF 0.35) — ER+ BC enriched
  • TP53 R248W missense (chr17:7675088, G→A, VAF 0.15) — subclonal

Patient-prefixed CNV profile (PAT002_cnv.cns):

  • chr13 BRCA2 loss (log2 −1.1, cn=1) — germline LOH
  • chr8 MYC gain (log2 +0.9, cn=3)
  • chr11 CCND1 gain (log2 +1.2, cn=3)
  • chr3 PIK3CA gain (log2 +0.7, cn=3)
  • chr16 CDH1 loss (log2 −0.8, cn=1)
  • chr17 ERBB2 neutral (log2 ~0, cn=2) — confirms HER2-negative status
  • 25 total segments spanning all autosomes + chrX

3. Multi-Omics Data (multiomics/)

Demonstration Data (Pre/Post Treatment Comparison)

File Size Description Source
tumor_rna_seq.csv 310 KB RNA-seq for 12 samples (6 pre, 6 post) Synthetic
tumor_proteomics.csv 125 KB Proteomics for 12 samples Synthetic
tumor_phosphoproteomics.csv 82 KB Phosphoproteomics for 12 samples Synthetic
sample_metadata.csv 0.5 KB Sample annotations (pre/post treatment) Synthetic

Demonstration specifications:

  • 12 samples: 6 pre-treatment biopsies + 6 post-treatment biopsies
  • All samples show treatment response (patient is disease-free)
  • Pre-processed counts matrices (not raw files)
  • Total size: ~518 KB

Generation method:

  • Random sampling from log-normal distributions
  • Pre-treatment: High proliferation (MKI67, PCNA, TOP2A), high ER/PR
  • Post-treatment: Low proliferation, slightly reduced ER/PR (tamoxifen), increased immune markers
  • Differential expression: 120 genes (12%), fold-changes 1.5-3.0×
  • BRCA2 expression reduced 50% (heterozygous germline mutation)
  • PI3K/AKT pathway activation (PIK3CA H1047R mutation effect)

Simulated biology:

  • RNA-seq: 1,000 genes (breast cancer markers + background)
    • High: ESR1, PGR, GATA3, FOXA1 (Luminal subtype)
    • Post-treatment ↓: MKI67, PCNA, proliferation markers
    • Post-treatment ↑: CD8A, CD3D, immune infiltration (TILs)
  • Proteomics: 400 proteins (ER, PR, HER2, Ki67, p53, AKT, mTOR)
  • Phosphoproteomics: 250 phosphosites (ESR1_S118, AKT1_S473, MTOR_S2448)
    • Post-treatment ↓: PI3K/AKT signaling

4. Spatial Transcriptomics (spatial/)

Demonstration Data (Simplified Visium)

File Size Description Source
visium_gene_expression.csv 190 KB Gene expression (35 genes × 900 spots) Synthetic
visium_spatial_coordinates.csv 35 KB Spatial coordinates for 900 spots Synthetic
visium_region_annotations.csv 19 KB Tissue region annotations (7 regions) Synthetic
filtered/visium_gene_expression_filtered.csv 178 KB QC-filtered expression matrix Synthetic
PAT002_expression.csv 187 KB Patient-prefixed expression (PAT002_ barcodes, luminal A profile) Synthetic
PAT002_coordinates.csv 45 KB Patient-prefixed spatial coordinates with patient_id column Synthetic
PAT002_regions.csv 38 KB Patient-prefixed regions (5 BC tissue types) Synthetic

Demonstration specifications:

  • 900 spots (30×30 grid, hexagonal layout)
  • 35 curated breast cancer genes
  • Total size: ~422 KB

Generation method:

  • Hexagonal Visium array layout (900 spots, 95% in-tissue)
  • 35 curated breast cancer-relevant genes
  • Spatial patterns using regional assignments:
    • Tumor core: High ER/PR, low proliferation (Luminal phenotype)
    • Proliferative zones: Ki67-high hotspots
    • Invasive front: Tumor-stroma boundary, EMT markers
    • Stroma: CAFs, collagen (COL1A1, ACTA2, FAP)
    • Immune zones: TILs (CD8A, CD3D, CD4)
    • Adipose: Breast adipose tissue (ADIPOQ, FABP4, LEP)
    • Normal ductal: Adjacent normal epithelium

Tissue regions simulated (7 breast-specific):

  1. tumor_core_luminal - ER+/PR+ tumor cells, low proliferation (7.4%)
  2. tumor_proliferative - Ki67-high tumor zone (4.7%)
  3. tumor_invasive_front - Invasion boundary (21.8%)
  4. stroma_fibrous - Dense collagen, CAFs (20.7%)
  5. stroma_immune - Lymphocyte infiltration (TILs) (13.1%)
  6. adipose - Normal breast adipose tissue (31.9%)
  7. ductal_normal - Adjacent normal ductal epithelium (0.4%)

Genes profiled (35 breast cancer markers):

  • Hormone receptors: ESR1, PGR, GATA3, FOXA1
  • Proliferation: MKI67, PCNA, TOP2A, CCND1, MYC
  • Tumor suppressors: BRCA2, GATA3, MAP3K1
  • Oncogenes: PIK3CA, AKT1, MTOR
  • Luminal markers: KRT8, KRT18, KRT19, EPCAM
  • Basal markers: KRT5, KRT14, EGFR (low expression)
  • Stromal: COL1A1, COL3A1, ACTA2, FAP, VIM
  • Immune: CD3D, CD8A, CD4, FOXP3, CD68, CD163
  • Hypoxia: HIF1A, VEGFA, CA9

5. Imaging Data (imaging/)

Directory Files Size Description Source
Main 7 TIFFs ~12 MB 1024×1024 immunofluorescence + H&E Synthetic microscopy
he/ 3 TIFFs ~25 MB H&E high/low resolution (2048×2048, 512×512) Synthetic histology
if/ 2 TIFFs ~16 MB ER/Ki67 high-resolution (2048×2048) Synthetic IF

Main directory (1024×1024, 16-bit grayscale):

  • PAT002_tumor_HE_20x.tiff - H&E histology (RGB 8-bit, pink/purple staining)
  • PAT002_tumor_IF_ER.tiff - Estrogen receptor (85% positive, nuclear)
  • PAT002_tumor_IF_PR.tiff - Progesterone receptor (70% positive, nuclear)
  • PAT002_tumor_IF_HER2.tiff - HER2 (negative, IHC 0, weak membrane)
  • PAT002_tumor_IF_KI67.tiff - Ki67 proliferation (20% positive, nuclear)
  • PAT002_tumor_IF_CD8.tiff - CD8+ T cells (8% positive, membrane)
  • PAT002_tumor_IF_DAPI.tiff - Nuclear staining (all cells)

High-resolution subdirectories:

  • he/PAT002_tumor_HE_20x_high.tif - 2048×2048 H&E
  • he/sample_002_high.tif, he/sample_002_low.tif - Additional H&E images
  • if/PAT002_tumor_IF_ER_high.tif - High-resolution ER staining
  • if/PAT002_tumor_IF_KI67_high.tif - High-resolution Ki67

Generation method:

  • Synthetic nuclei: 150-450 elliptical cells per image (size-dependent)
  • Nuclear markers (ER/PR/Ki67): Circular gradients in nucleus
  • Membrane markers (HER2/CD8): Ring patterns around cells
  • H&E simulation: Pink eosin (cytoplasm/stroma) + purple hematoxylin (nuclei)
  • Gaussian noise (σ=200-250), PSF blur (σ=1.2-1.5)
  • 16-bit grayscale (0-65535 range), RGB for H&E

Marker specifications:

  • ER: 85% cells positive, strong nuclear signal (9,000-15,000 intensity)
  • PR: 70% cells positive, moderate nuclear signal
  • HER2: Negative (all cells weak membrane, 800-2,000 intensity)
  • Ki67: 20% cells positive (post-treatment proliferation)
  • CD8: 8% cells positive (tumor-infiltrating lymphocytes)
  • H&E: Realistic histology with nuclear detail and stromal texture

6. h5ad Files (AnnData Format)

Spatial Single-Cell Data (quantum/)

File Size Description Source
PAT002_tumor_spatial.h5ad 7.9 MB 500 cells × 2,000 genes, spatial coordinates Synthetic AnnData

Contents:

  • Cells: 500 single cells with spatial coordinates (x, y)
  • Genes: 2,000 genes (79 breast cancer markers + 1,921 background)
  • Cell types (8): tumor_cells_luminal (36%), CAFs (20%), CD8_T_cells (14%), CD4_T_cells (10%), macrophages (8%), B_cells (6%), endothelial (4%), adipocytes (2%)
  • Spatial organization:
    • Tumor core at center (luminal subtype, high ER/PR)
    • CAFs surrounding tumor
    • TILs at margins (CD8+, CD4+)
    • Adipocytes at periphery (breast-specific)
  • Metadata: patient_id, diagnosis, receptor_status, BRCA2_status, treatment_status

Perturbation Data (perturbation/)

File Size Description Source
pat002_tcells.h5ad 463 KB 500 cells × 100 genes, CRISPR screen Synthetic perturbation
gears_pat002_results.json 2.1 KB GEARS GNN perturbation predictions (5 perturbations) Synthetic

GEARS Predictions (gears_pat002_results.json):

  • ESR1 knockout — loss of luminal identity, models tamoxifen resistance (confidence 0.91)
  • PIK3CA inhibition — reduced proliferation, alpelisib sensitivity prediction (confidence 0.87)
  • CDK4 inhibition — G1 arrest, palbociclib response prediction (confidence 0.93)
  • BRCA2 knockout — HR deficiency, PARP inhibitor sensitivity (confidence 0.89)
  • FOXP3 overexpression — immune suppression, Treg expansion limiting immunotherapy (confidence 0.82)
  • Summary: most actionable target CDK4, HRD score 35, immune evasion 0.41, luminal stability 0.78

Contents:

  • Experiment: PD-1 (PDCD1) knockout CRISPR screen in CD8+ T cells
  • Conditions: Control (250 cells) vs PDCD1_KO (250 cells)
  • Genes: 100 T cell markers (effector, exhaustion, proliferation, metabolism)
  • Expected phenotype:
    • Control: High PD-1 (PDCD1=339), high exhaustion (TOX=161)
    • PDCD1 KO: No PD-1 (PDCD1=0.8), restored effector function (GZMB=203, IFNG=206, 4.4× FC)
    • PDCD1 KO: Increased proliferation (MKI67=91, 5.6× FC)
  • Purpose: Demonstrate immune checkpoint blockade effects for BRCA2+ breast cancer immunotherapy

Data Quality & Realism

Realistic Features ✅

  1. Clinically plausible genetics - BRCA2 c.5946delT pathogenic variant + PIK3CA H1047R hotspot
  2. Biologically coherent - ER+/PR+ Luminal phenotype with HRD signature
  3. Spatial organization - Breast tumor microenvironment with adipocytes, CAFs, TILs
  4. Temporal dynamics - Pre/post treatment comparison showing therapy response
  5. Treatment trajectory - Realistic adjuvant therapy sequence (AC-T chemo + radiation + tamoxifen)
  6. Surveillance pattern - Normal CEA/CA 15-3 trajectory, disease-free status

Synthetic Limitations ❌

  1. No batch effects - Real data would have technical variation across timepoints/platforms
  2. Simplified biology - Real tumors have >20,000 genes, not 35-1,000
  3. Idealized spatial patterns - Real tissue is more heterogeneous, with necrosis and artifacts
  4. No sequencing errors - Real VCF would have mapping quality issues, strand bias
  5. Perfect correlations - Real multi-omics has more noise and missing data
  6. Simplified family history - Real BRCA2 families have more complex pedigrees

Intended Use Cases

✅ Appropriate Uses

  • Workflow development - Test MCP server integration for breast cancer precision medicine
  • CI/CD testing - Automated pipeline validation (genomics, imaging, spatial)
  • Demonstration - Show capabilities to stakeholders (oncologists, bioinformaticians)
  • Education - Teach breast cancer biology, BRCA2 testing, HRD, immunotherapy
  • Software testing - Verify analysis algorithms (ER/PR/HER2 scoring, spatial deconvolution)
  • FHIR implementation - Test HL7 FHIR R4 interoperability

❌ Inappropriate Uses

  • Clinical decision-making - NOT real patient data, do not use for treatment decisions
  • Research publications - Cannot be cited as real clinical data
  • Regulatory submissions - Not acceptable for FDA/EMA applications
  • IRB protocols - Not subject to human subjects research oversight
  • Training machine learning models - Synthetic data may not generalize to real patients

Real Data Alternatives

If you need real breast cancer patient data, consider these public resources:

  1. TCGA-BRCA (The Cancer Genome Atlas - Breast Cancer)

  2. 10x Genomics Public Datasets

  3. METABRIC (Molecular Taxonomy of Breast Cancer)

  4. ClinVar BRCA1/BRCA2 Variants

    • ftp://ftp.ncbi.nlm.nih.gov/pub/clinvar/vcf_GRCh38/
    • Clinical significance of BRCA germline variants
    • Pathogenic/benign classification

  5. BRCA Exchange

  6. Single Cell Portal (Broad Institute)

Technical Details

File Formats

Format Usage Tools
JSON Clinical data, FHIR resources jq, Python json
VCF Genomic variants bcftools, vcftools, GATK
CSV Multi-omics matrices, spatial data Python pandas, R read.csv
TIFF Microscopy images (8/16-bit) ImageJ/Fiji, scikit-image, QuPath
h5ad AnnData (single-cell, spatial) Python scanpy, anndata

Dataset Size

Demonstration (current):

  • Clinical: ~50 KB (FHIR + JSON)
  • Genomics: ~6 KB (VCFs + CNS)
  • Multi-omics: ~520 KB (CSV matrices)
  • Spatial: ~695 KB (Visium CSVs + patient-prefixed CSVs)
  • Imaging: ~55 MB (TIFFs)
  • h5ad: ~8.4 MB (AnnData)
  • Perturbation: ~465 KB (h5ad + GEARS JSON)
  • Total: ~65 MB (demonstration-scale)

Production (typical hospital per-patient):

  • Clinical: ~500 KB (full EHR extract)
  • Genomics: ~3 GB (WGS BAM files), ~10 MB (VCF)
  • Multi-omics: ~5 GB (RNA-seq BAM), ~500 MB (proteomics raw), ~20 MB (matrices)
  • Spatial: ~500 MB (Visium H5 matrix + histology image)
  • Imaging: ~1-5 GB (whole slide images, multiplex IF)
  • Total: ~10-15 GB per patient (raw files), ~100-200 MB (processed matrices)

Data License

This synthetic dataset is released under the Apache License 2.0.

Copyright 2026 Lynn Langit

Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at

    http://www.apache.org/licenses/LICENSE-2.0

Attribution

If using this synthetic dataset in presentations or educational materials, please attribute:

Synthetic patient data (PAT002-BC-2026) generated for MCP server demonstration
Source: github.com/lynnlangit/precision-medicine-mcp
License: Apache 2.0

Generation Scripts

All data in this directory was generated using Python scripts:

Script Purpose Location
(manual creation) Clinical FHIR JSON files clinical/fhir_raw/
(manual creation) Genomics VCF file genomics/
generate_PAT002_multiomics.py Multi-omics CSV matrices multiomics/
generate_PAT002_spatial.py Spatial transcriptomics CSVs spatial/
generate_PAT002_imaging.py Microscopy TIFF images imaging/
create_PAT002_spatial_data.py Spatial h5ad file quantum/
generate_PAT002_perturbation.py Perturbation h5ad file perturbation/
generate_PAT002_missing_files.py 6 patient-prefixed files (CNV, somatic VCF, spatial CSVs, GEARS JSON) Root

Dependencies:

  • Python 3.9+
  • NumPy, Pandas, SciPy
  • scikit-image, Pillow (imaging)
  • anndata, scanpy (h5ad)

To regenerate data:

# Multi-omics
python3 multiomics/generate_PAT002_multiomics.py

# Spatial transcriptomics
python3 spatial/generate_PAT002_spatial.py

# Imaging
python3 imaging/generate_PAT002_imaging.py

# h5ad files
python3 quantum/create_PAT002_spatial_data.py
python3 perturbation/generate_PAT002_perturbation.py

# Patient-prefixed files (CNV, somatic VCF, spatial CSVs, GEARS JSON)
python3 generate_PAT002_missing_files.py

Patient Summary

Michelle Anne Thompson (synthetic, 42 years old, female)

Medical History:

  • Diagnosed December 2024 with Stage IIA (T2N0M0) ER+/PR+/HER2- Invasive Ductal Carcinoma, Left Breast
  • BRCA2 germline pathogenic variant (c.5946delT, frameshift, Class 5)
  • Strong family history: Mother breast cancer (age 58), maternal aunt ovarian cancer (age 52), sister BRCA2+ carrier (no cancer)

Treatment Timeline:

  • January 2025: Lumpectomy + sentinel lymph node biopsy (node-negative)
  • February-April 2025: AC chemotherapy (doxorubicin 60mg/m² + cyclophosphamide q21d × 4 cycles)
  • April-May 2025: Paclitaxel 80mg/m² IV weekly × 12 weeks
  • June-July 2025: Adjuvant radiation therapy (whole breast + tumor bed boost)
  • August 2025: Started tamoxifen 20mg PO daily (planned 5-10 years)

Current Status (January 2026):

  • Disease-free (No Evidence of Disease, NED)
  • CEA 2.1 ng/mL (normal), CA 15-3 20 U/mL (normal)
  • Tolerating tamoxifen well (mild hypercholesterolemia, expected side effect)
  • Surveillance: Q3 months with tumor markers, annual mammography

Genomic Profile:

  • BRCA2 c.5946delT heterozygous (germline, pathogenic)
  • PIK3CA H1047R (somatic, hotspot)
  • MYC amplification, CCND1 amplification
  • TP53 wild-type (favorable for Luminal subtype)
  • ESR1 wild-type (tamoxifen-sensitive)

Clinical Significance:

  • PARP inhibitor eligible (BRCA2+, HRD phenotype)
  • Good prognosis (Stage IIA, node-negative, ER+/PR+, responsive to therapy)
  • Surveillance for second primary breast/ovarian cancer (BRCA2 carrier)
  • Family counseling recommended (genetic testing for siblings/children)

Last Updated: 2026-05-06 | Status: 100% Synthetic — Research/Demo Only