Install any skill in seconds. Free to start, no credit card required.
Get Started Free →AI-powered homologous recombination deficiency (HRD) analysis for PARP inhibitor response prediction using genomic scarring signatures and BRCA pathway assessment.
.claude/skills/hrd-analysis-agent/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-03 | ✗→✓ | ▲ Improved | 9% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 46% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 66% | 0% |
| case-14 | ✗→✓ | ▲ Improved | 72% | 0% |
| case-18 | ✗→✓ | ▲ Improved | 7% | 0% |
<!--
#
#
-->
The HRD Analysis Agent provides comprehensive analysis of homologous recombination deficiency for predicting response to PARP inhibitors and platinum chemotherapy. It integrates genomic scarring signatures (LOH, TAI, LST), BRCA1/2 status, and HRD gene pathway analysis.
| Score | Definition | Biological Basis | |-------|------------|------------------| | LOH | Loss of heterozygosity regions >15 Mb | Genomic scarring | | TAI | Telomeric allelic imbalance | End-to-end fusions | | LST | Large-scale state transitions | Break-induced repair | | GIS | Combined LOH + TAI + LST | Overall HRD score |
HRD-Positive Threshold: GIS ≥ 42 (Myriad myChoice) or equivalent
| Gene | Function | HRD Contribution | |------|----------|------------------| | BRCA1 | HR core component | Major | | BRCA2 | RAD51 loading | Major | | PALB2 | BRCA2 partner | Moderate-Major | | RAD51C/D | HR mediators | Moderate | | ATM | DNA damage sensing | Moderate | | CHEK2 | Cell cycle checkpoint | Moderate | | BARD1 | BRCA1 partner | Moderate | | BRIP1 | Fanconi pathway | Moderate | | CDK12 | Transcription | Variable | | RAD51B | HR mediator | Low-Moderate |
User: "Analyze HRD status for this ovarian cancer patient to guide PARP inhibitor selection."
Agent Action:
bashpython3 Skills/Oncology/HRD_Analysis_Agent/hrd_analyzer.py \ --cnv_segments tumor_segments.tsv \ --mutations somatic_variants.maf \ --germline germline_variants.vcf \ --tumor_type ovarian \ --purity 0.65 \ --ploidy 2.1 \ --output hrd_report.json
| Test | Components | Threshold | FDA Status | |------|------------|-----------|------------| | myChoice CDx | GIS + BRCA | ≥42 or BRCA+ | FDA approved | | FoundationOne | LOH | ≥16% | FDA approved | | SOPHiA DDM HRD | GIS + BRCA | ≥42 or BRCA+ | CE-IVD |
FDA-Approved PARP Inhibitor Indications:
| Status | PARP Response | Platinum Response | |--------|---------------|-------------------| | BRCA mutated | Very high | High | | HRD+ / BRCA WT | High | Moderate-High | | HRD- / BRCA WT | Limited | Standard | | Reversion+ | Poor | Poor |
GIS Calculation:
Reversion Detection:
Response Prediction:
| Mechanism | Detection | Implication | |-----------|-----------|-------------| | BRCA reversion | Sequencing | Acquired resistance | | 53BP1 loss | Expression/mutation | Rescued HR | | ABCB1 upregulation | Expression | Drug efflux | | PARP1 loss | Expression | Target loss |
AI Group - Biomedical AI Platform
<!-- AUTHOR_SIGNATURE: 9a7f3c2e-MD-BABU-MIA-2026-MSSM-SECURE -->
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-06 | pass→pass | 10,247 | 4,592 | -55% | 1 | 1 | 0% | 1,941 | 2,240 | +15% | 0 | 0 | — |
case-15 | fail→fail | 11,979 | 9,361 | -22% | 1 | 1 | 0% | 2,138 | 3,140 | +47% | 0 | 0 | — |
case-01 | pass→pass | 6,481 | 4,807 | -26% | 1 | 1 | 0% | 1,305 | 2,389 | +83% | 0 | 0 | — |
case-02 | pass→pass | 7,631 | 6,121 | -20% | 1 | 1 | 0% | 1,442 | 2,730 | +89% | 0 | 0 | — |
case-03 | fail→pass | 21,101 | 3,685 | -83% | 1 | 1 | 0% | 1,987 | 2,168 | +9% | 0 | 0 | — |
case-04 | pass→pass | 8,596 | 10,617 | +24% | 1 | 1 | 0% | 1,501 | 3,432 | +129% | 0 | 0 | — |
case-05 | fail→pass | 10,289 | 7,407 | -28% | 1 | 1 | 0% | 1,888 | 2,754 | +46% | 0 | 0 | — |
case-07 | fail→pass | 11,149 | 10,265 | -8% | 1 | 1 | 0% | 1,963 | 3,258 | +66% | 0 | 0 | — |
case-08 | pass→pass | 7,648 | 5,603 | -27% | 1 | 1 | 0% | 1,357 | 2,536 | +87% | 0 | 0 | — |
case-09 | pass→pass | 8,290 | 5,609 | -32% | 1 | 1 | 0% | 1,445 | 2,432 | +68% | 0 | 0 | — |
case-10 | pass→pass | 11,069 | 10,545 | -5% | 1 | 1 | 0% | 2,396 | 3,588 | +50% | 0 | 0 | — |
case-11 | pass→pass | 4,870 | 5,260 | +8% | 1 | 1 | 0% | 918 | 2,404 | +162% | 0 | 0 | — |
case-12 | pass→pass | 10,484 | 10,261 | -2% | 1 | 1 | 0% | 1,854 | 3,300 | +78% | 0 | 0 | — |
case-13 | pass→pass | 5,156 | 6,799 | +32% | 1 | 1 | 0% | 934 | 2,590 | +177% | 0 | 0 | — |
case-14 | fail→pass | 8,678 | 7,293 | -16% | 1 | 1 | 0% | 1,644 | 2,828 | +72% | 0 | 0 | — |
case-16 | pass→pass | 6,312 | 3,833 | -39% | 1 | 1 | 0% | 1,011 | 2,039 | +102% | 0 | 0 | — |
case-17 | pass→pass | 11,347 | 5,178 | -54% | 1 | 1 | 0% | 1,989 | 2,358 | +19% | 0 | 0 | — |
case-18 | fail→pass | 10,775 | 3,552 | -67% | 1 | 1 | 0% | 1,886 | 2,018 | +7% | 0 | 0 | — |
case-19 | pass→pass | 13,442 | 12,699 | -6% | 1 | 1 | 0% | 2,314 | 3,551 | +53% | 0 | 0 | — |
case-20 | pass→pass | 5,886 | 4,297 | -27% | 1 | 1 | 0% | 1,225 | 2,329 | +90% | 0 | 0 | — |
case-21 | pass→pass | 5,456 | 5,071 | -7% | 1 | 1 | 0% | 1,109 | 2,463 | +122% | 0 | 0 | — |
case-22 | fail→fail | 7,039 | 11,974 | +70% | 1 | 1 | 0% | 1,497 | 3,965 | +165% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted, and 21 counted toward the lift figure. The other 1 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +23 percentage points is the difference between those two pass rates over the 21 comparable cases.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
| Model | Method | Date | Lift |
|---|---|---|---|
| gemini-3.6-flash | verified | 7/26/2026 | +5% |
Other measured skills in the registry, with their headline benchmark lift.