Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →AI-driven integration of cellular imaging, laser microdissection, and ultra-sensitive mass spectrometry for spatially-resolved single-cell proteomics.
.claude/skills/deep-visual-proteomics-agent/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | -27% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 12% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -12% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 10% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 7% | 0% |
<!--
#
#
-->
The Deep Visual Proteomics Agent implements the Deep Visual Proteomics (DVP) workflow that combines AI-driven image analysis of cellular phenotypes with automated laser microdissection and ultra-high-sensitivity mass spectrometry. It links protein abundance to complex cellular or subcellular phenotypes while preserving spatial context.
Tissue Section
↓
[AI Image Analysis] → Cell Segmentation → Phenotype Classification
↓
[Region Selection] → LMD Coordinates → Automated Microdissection
↓
[Sample Processing] → Low-input LC-MS/MS → Proteome Quantification
↓
[Data Integration] → Spatial Proteome Map → Pathway AnalysisUser: "Identify tumor vs. stroma cells in this H&E image and generate proteome profiles for each population."
Agent Action:
bashpython3 Skills/Proteomics/Deep_Visual_Proteomics_Agent/dvp_analyzer.py \ --image tissue_section.tiff \ --segmentation cellpose \ --classifier tumor_stroma_cnn \ --generate_lmd true \ --ms_data maxquant_output/ \ --analysis differential \ --output dvp_results/
| Component | Tool/Method | Description | |-----------|-------------|-------------| | Segmentation | Cellpose, StarDist | Instance segmentation of cells | | Classification | Custom CNN/ViT | Phenotype assignment | | LMD Interface | Leica LMD7, PALM | Coordinate export formats | | MS Processing | MaxQuant, DIA-NN | Protein quantification | | Integration | Custom Python | Spatial mapping |
The agent implements BINNs that integrate:
Input: Protein abundances
↓
Pathway Layer: Proteins → Pathways (sparse connections)
↓
Process Layer: Pathways → Biological processes
↓
Output: Phenotype classification + pathway importance scoresSensitivity: 100-500 cells per sample for robust quantification Throughput: 1,000-5,000 proteins per sample Resolution: Single-cell to ~10-cell resolution Formats: TIFF/SVS images, MaxQuant/DIA-NN output
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-21 | pass→pass | 20,309 | 21,037 | +4% | 1 | 1 | 0% | 3,585 | 5,574 | +55% | 0 | 0 | — |
case-22 | pass→pass | 15,949 | 11,436 | -28% | 1 | 1 | 0% | 3,048 | 3,479 | +14% | 0 | 0 | — |
case-01 | fail→fail | 19,921 | 5,422 | -73% | 1 | 1 | 0% | 3,779 | 1,331 | -65% | 0 | 0 | — |
case-11 | pass→pass | 11,475 | 1,850 | -84% | 1 | 1 | 0% | 1,988 | 1,427 | -28% | 0 | 0 | — |
case-02 | fail→fail | 16,572 | 28,820 | +74% | 1 | 1 | 0% | 3,183 | 7,271 | +128% | 0 | 0 | — |
case-03 | fail→fail | 18,372 | 6,459 | -65% | 1 | 1 | 0% | 2,950 | 1,500 | -49% | 0 | 0 | — |
case-04 | fail→pass | 13,263 | 2,957 | -78% | 1 | 1 | 0% | 2,211 | 1,619 | -27% | 0 | 0 | — |
case-05 | fail→pass | 7,672 | 2,171 | -72% | 1 | 1 | 0% | 1,368 | 1,536 | +12% | 0 | 0 | — |
case-20 | pass→pass | 17,972 | 14,014 | -22% | 1 | 1 | 0% | 3,627 | 3,879 | +7% | 0 | 0 | — |
case-06 | fail→pass | 15,030 | 6,408 | -57% | 1 | 1 | 0% | 2,808 | 2,460 | -12% | 0 | 0 | — |
case-07 | fail→pass | 7,917 | 2,222 | -72% | 1 | 1 | 0% | 1,381 | 1,524 | +10% | 0 | 0 | — |
case-08 | fail→pass | 8,495 | 2,477 | -71% | 1 | 1 | 0% | 1,441 | 1,542 | +7% | 0 | 0 | — |
case-09 | fail→pass | 6,376 | 2,589 | -59% | 1 | 1 | 0% | 1,055 | 1,489 | +41% | 0 | 0 | — |
case-10 | pass→pass | 18,490 | 10,238 | -45% | 1 | 1 | 0% | 3,064 | 2,906 | -5% | 0 | 0 | — |
case-12 | fail→fail | 12,888 | 5,981 | -54% | 1 | 1 | 0% | 2,046 | 2,035 | -1% | 0 | 0 | — |
case-13 | fail→pass | 18,830 | 3,294 | -83% | 1 | 1 | 0% | 1,258 | 1,671 | +33% | 0 | 0 | — |
case-14 | fail→pass | 20,451 | 3,910 | -81% | 1 | 1 | 0% | 1,133 | 1,788 | +58% | 0 | 0 | — |
case-15 | fail→pass | 9,037 | 4,469 | -51% | 1 | 1 | 0% | 1,653 | 1,913 | +16% | 0 | 0 | — |
case-16 | pass→pass | 9,489 | 7,869 | -17% | 1 | 1 | 0% | 1,567 | 2,331 | +49% | 0 | 0 | — |
case-17 | pass→pass | 7,247 | 1,440 | -80% | 1 | 1 | 0% | 1,187 | 1,363 | +15% | 0 | 0 | — |
case-18 | pass→pass | 15,867 | 8,369 | -47% | 1 | 1 | 0% | 2,587 | 2,460 | -5% | 0 | 0 | — |
case-19 | fail→pass | 10,128 | 1,327 | -87% | 1 | 1 | 0% | 1,584 | 1,291 | -18% | 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 18 counted toward the lift figure. The other 4 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 +45 percentage points is the difference between those two pass rates over the 18 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 | 7/24/2026 | +14% |
Other measured skills in the registry, with their headline benchmark lift.