AgentSkillsCN

Skill

技能

SKILL.md

Personal Genomics Skill v4.4.0

Comprehensive local DNA analysis with 1600+ markers across 30 categories. Privacy-first genetic analysis for AI agents.

🆕 v4.4.0: 1000 GENOMES POPULATION COMPARISON & ANCIENT DNA

  • •Transparent population comparison (not black-box percentages)
  • •Ancient DNA signal detection (WHG, ANF, Yamnaya, Neanderthal)
  • •Interactive dashboard with population frequency visualizations
  • •Complete methodology documentation

⚠️ v4.3.0 focuses on ACCURACY AND HONESTY - improved uncertainty handling, PMIDs for all claims, and explicit limitations.

Quick Start

bash
python comprehensive_analysis.py /path/to/dna_file.txt

⚠️ Important Limitations

  1. •

    Haplogroups are LOW CONFIDENCE - Consumer arrays cannot reliably call haplogroups. Recommend dedicated Y-DNA/mtDNA testing (FTDNA, YFull) for accuracy.

  2. •

    Ancestry shows ANCIENT SIGNALS, not modern ethnicity - Modern ethnicity percentages are unreliable. Instead we detect signals from well-characterized ancient populations (WHG, Neolithic Farmers, Steppe, Neanderthal, Denisovan).

  3. •

    PRS scores show RANGES, not point estimates - Polygenic risk scores have wide confidence intervals. Most conditions are 50-80% non-genetic.

  4. •

    Every marker has PMIDs - All claims are backed by literature citations linked to PubMed.

Triggers

Activate this skill when user mentions:

  • •DNA analysis, genetic analysis, genome analysis
  • •23andMe, AncestryDNA, MyHeritage results
  • •Pharmacogenomics, drug-gene interactions
  • •Medication interactions, drug safety
  • •Genetic risk, disease risk, health risk
  • •Carrier status, carrier testing
  • •VCF file analysis
  • •APOE, MTHFR, CYP2D6, BRCA, or other gene names
  • •Polygenic risk scores
  • •Haplogroups, maternal lineage, paternal lineage
  • •Ancestry composition, ethnicity
  • •Hereditary cancer, Lynch syndrome
  • •Autoimmune genetics, HLA, celiac
  • •Pain sensitivity, opioid response
  • •Sleep optimization, chronotype, caffeine metabolism
  • •Dietary genetics, lactose intolerance, celiac
  • •Athletic genetics, sports performance
  • •UV sensitivity, skin type, melanoma risk
  • •Telomere length, longevity genetics

Supported Files

  • •23andMe, AncestryDNA, MyHeritage, FTDNA
  • •VCF files (whole genome/exome, .vcf or .vcf.gz)
  • •Any tab-delimited rsid format

Output Location

~/dna-analysis/reports/

  • •agent_summary.json - AI-optimized, priority-sorted
  • •full_analysis.json - Complete data
  • •report.txt - Human-readable
  • •genetic_report.pdf - Professional PDF report
  • •dashboard.html - Interactive visualization

New v4.3.0 Features (Accuracy Update)

Honest Haplogroup Reporting

  • •LOW CONFIDENCE labels on all haplogroup calls
  • •Explicit disclaimer that consumer arrays can't reliably call haplogroups
  • •Recommendations for dedicated Y-DNA/mtDNA testing services
  • •PMIDs for haplogroup marker sources

Ancient Ancestral Signals (Replaces Modern Ethnicity)

  • •Western Hunter-Gatherers (WHG) - Mesolithic Europeans (~15,000-8,000 BP)
  • •Early European Farmers (EEF) - Neolithic Anatolians (~10,000-5,000 BP)
  • •Steppe Pastoralists - Yamnaya/Bronze Age (~5,000-4,000 BP)
  • •Neanderthal Introgression - Archaic human (~50,000-40,000 BP)
  • •Denisovan Introgression - Archaic human (high-altitude adaptation)
  • •Shows "Signals Detected" not percentages
  • •Includes time periods and trait contributions
  • •Based on ancient DNA studies with PMIDs

PRS with Uncertainty Ranges

  • •Percentile RANGES instead of point estimates
  • •Confidence intervals based on marker coverage
  • •Explicit interpretation guidance ("likely average", "uncertain", etc.)

PMIDs Throughout

  • •Every marker has at least one literature citation
  • •Clickable PubMed links in dashboard
  • •Methodology & Limitations section in dashboard

Legacy v4.0-4.2 Features

Haplogroup Analysis (indicative only)

  • •Mitochondrial DNA (mtDNA) - maternal lineage
  • •Y-chromosome - paternal lineage (males only)
  • •Migration history context
  • •PhyloTree/ISOGG standards

Ancient Ancestry (scientifically grounded)

  • •Detection of ancient population signals
  • •Based on well-characterized ancient DNA
  • •Includes archaic introgression (Neanderthal/Denisovan)

Hereditary Cancer Panel

  • •BRCA1/BRCA2 comprehensive
  • •Lynch syndrome (MLH1, MSH2, MSH6, PMS2)
  • •Other genes (APC, TP53, CHEK2, PALB2, ATM)
  • •ACMG-style classification

Autoimmune HLA

  • •Celiac (DQ2/DQ8) - can rule out if negative
  • •Type 1 Diabetes
  • •Ankylosing spondylitis (HLA-B27)
  • •Rheumatoid arthritis, lupus, MS

Pain Sensitivity

  • •COMT Val158Met
  • •OPRM1 opioid receptor
  • •SCN9A pain signaling
  • •TRPV1 capsaicin sensitivity
  • •Migraine susceptibility

PDF Reports

  • •Professional format
  • •Physician-shareable
  • •Executive summary
  • •Detailed findings
  • •Disclaimers included

New v4.1.0 Features

Medication Interaction Checker

python
from markers.medication_interactions import check_medication_interactions

result = check_medication_interactions(
    medications=["warfarin", "clopidogrel", "omeprazole"],
    genotypes=user_genotypes
)
# Returns critical/serious/moderate interactions with alternatives
  • •Accepts brand or generic names
  • •CPIC guidelines integrated
  • •PubMed citations included
  • •FDA warning flags

Sleep Optimization Profile

python
from markers.sleep_optimization import generate_sleep_profile

profile = generate_sleep_profile(genotypes)
# Returns ideal wake/sleep times, coffee cutoff, etc.
  • •Chronotype (morning/evening preference)
  • •Caffeine metabolism speed
  • •Personalized timing recommendations

Dietary Interaction Matrix

python
from markers.dietary_interactions import analyze_dietary_interactions

diet = analyze_dietary_interactions(genotypes)
# Returns food-specific guidance
  • •Caffeine, alcohol, saturated fat, lactose, gluten
  • •APOE-specific diet recommendations
  • •Bitter taste perception

Athletic Performance Profile

python
from markers.athletic_profile import calculate_athletic_profile

profile = calculate_athletic_profile(genotypes)
# Returns power/endurance type, recovery profile, injury risk
  • •Sport suitability scoring
  • •Training recommendations
  • •Injury prevention guidance

UV Sensitivity Calculator

python
from markers.uv_sensitivity import generate_uv_sensitivity_report

uv = generate_uv_sensitivity_report(genotypes)
# Returns skin type, SPF recommendation, melanoma risk
  • •Fitzpatrick skin type estimation
  • •Vitamin D synthesis capacity
  • •Melanoma risk factors

Natural Language Explanations

python
from markers.explanations import generate_plain_english_explanation

explanation = generate_plain_english_explanation(
    rsid="rs3892097", gene="CYP2D6", genotype="GA",
    trait="Drug metabolism", finding="Poor metabolizer carrier"
)
  • •Plain-English summaries
  • •Research variant flagging
  • •PubMed links

Telomere & Longevity

python
from markers.advanced_genetics import estimate_telomere_length

telomere = estimate_telomere_length(genotypes)
# Returns relative estimate with appropriate caveats
  • •TERT, TERC, OBFC1 variants
  • •Longevity associations (FOXO3, APOE)

Data Quality

  • •Call rate analysis
  • •Platform detection
  • •Confidence scoring
  • •Quality warnings

Export Formats

  • •Genetic counselor clinical export
  • •Apple Health compatible
  • •API-ready JSON
  • •Integration hooks

Marker Categories (21 total)

  1. •Pharmacogenomics (159) - Drug metabolism
  2. •Polygenic Risk Scores (277) - Disease risk
  3. •Carrier Status (181) - Recessive carriers
  4. •Health Risks (233) - Disease susceptibility
  5. •Traits (163) - Physical/behavioral
  6. •Haplogroups (44) - Lineage markers
  7. •Ancestry (124) - Population informative
  8. •Hereditary Cancer (41) - BRCA, Lynch, etc.
  9. •Autoimmune HLA (31) - HLA associations
  10. •Pain Sensitivity (20) - Pain/opioid response
  11. •Rare Diseases (29) - Rare conditions
  12. •Mental Health (25) - Psychiatric genetics
  13. •Dermatology (37) - Skin and hair
  14. •Vision & Hearing (33) - Sensory genetics
  15. •Fertility (31) - Reproductive health
  16. •Nutrition (34) - Nutrigenomics
  17. •Fitness (30) - Athletic performance
  18. •Neurogenetics (28) - Cognition/behavior
  19. •Longevity (30) - Aging markers
  20. •Immunity (43) - HLA and immune
  21. •Ancestry AIMs (24) - Admixture markers

Agent Integration

The agent_summary.json provides:

json
{
  "critical_alerts": [],
  "high_priority": [],
  "medium_priority": [],
  "pharmacogenomics_alerts": [],
  "apoe_status": {},
  "polygenic_risk_scores": {},
  "haplogroups": {
    "mtDNA": {"haplogroup": "H", "lineage": "maternal"},
    "Y_DNA": {"haplogroup": "R1b", "lineage": "paternal"}
  },
  "ancestry": {
    "composition": {},
    "admixture": {}
  },
  "hereditary_cancer": {},
  "autoimmune_risk": {},
  "pain_sensitivity": {},
  "lifestyle_recommendations": {
    "diet": [],
    "exercise": [],
    "supplements": [],
    "avoid": []
  },
  "drug_interaction_matrix": {},
  "data_quality": {}
}

Critical Findings (Always Alert User)

Pharmacogenomics

  • •DPYD variants - 5-FU/capecitabine FATAL toxicity risk
  • •HLA-B*5701 - Abacavir hypersensitivity
  • •HLA-B*1502 - Carbamazepine SJS (certain populations)
  • •MT-RNR1 - Aminoglycoside-induced deafness

Hereditary Cancer

  • •BRCA1/BRCA2 pathogenic - Breast/ovarian cancer syndrome
  • •Lynch syndrome genes - Colorectal/endometrial cancer
  • •TP53 pathogenic - Li-Fraumeni syndrome (multi-cancer)

Disease Risk

  • •APOE ε4/ε4 - ~12x Alzheimer's risk
  • •Factor V Leiden - Thrombosis risk, contraceptive implications
  • •HLA-B27 - Ankylosing spondylitis susceptibility (OR ~70)

Carrier Status

  • •CFTR - Cystic fibrosis (1 in 25 Europeans)
  • •HBB - Sickle cell (1 in 12 African Americans)
  • •HEXA - Tay-Sachs (1 in 30 Ashkenazi Jews)

Usage Examples

Basic Analysis

python
from comprehensive_analysis import main
main()  # Uses command line args

Haplogroup Analysis

python
from markers.haplogroups import analyze_haplogroups
result = analyze_haplogroups(genotypes)
print(result["mtDNA"]["haplogroup"])  # e.g., "H"

Ancestry

python
from markers.ancestry_composition import get_ancestry_summary
ancestry = get_ancestry_summary(genotypes)

Cancer Panel

python
from markers.cancer_panel import analyze_cancer_panel
cancer = analyze_cancer_panel(genotypes)
if cancer["pathogenic_variants"]:
    print("ALERT: Pathogenic variants detected")

Generate PDF

python
from pdf_report import generate_pdf_report
pdf_path = generate_pdf_report(analysis_results)

Export for Genetic Counselor

python
from exports import generate_genetic_counselor_export
clinical = generate_genetic_counselor_export(results, "clinical.json")

Privacy

  • •All analysis runs locally
  • •Zero network requests
  • •No data leaves the machine

Limitations

  • •Consumer arrays miss rare variants (~0.1% of genome)
  • •Results are probabilistic, not deterministic
  • •Not a medical diagnosis
  • •Most conditions 50-80% non-genetic
  • •Consult healthcare providers for medical decisions
  • •Negative hereditary cancer result does NOT rule out cancer syndrome
  • •Haplogroup resolution limited without WGS

When to Recommend Genetic Counseling

  • •Any pathogenic hereditary cancer variant
  • •APOE ε4/ε4 genotype
  • •Multiple critical pharmacogenomic findings
  • •Carrier status with reproduction implications
  • •High-risk autoimmune HLA types with symptoms
  • •Results causing significant user distress