Skip to content

DEMO_AND_PITCH.md

Latest commit

 

History

History
108 lines (71 loc) · 3.3 KB

DEMO_AND_PITCH.md

File metadata and controls

108 lines (71 loc) · 3.3 KB

Demo & Pitch Guide

Copy-paste demo scripts for funders, grant reviewers, and hospital decision-makers. Uses synthetic PatientOne data (PAT001-OVC-2025) -- 100% safe, no PHI.

Prerequisites: Claude Desktop with MCP servers configured (setup guide), OR deployed Streamlit UI, OR Jupyter notebook.

90-Second Demo Script

Prompt 1: Show the Problem (15s)

What data modalities need to be integrated for a comprehensive Stage IV
ovarian cancer precision medicine analysis?

Talking point: "Each modality traditionally requires separate tools and 8+ hours."

Prompt 2: Show the Solution (30s)

Using PatientOne data (PAT001-OVC-2025), identify the top 3 actionable
treatment targets based on spatial transcriptomics pathway enrichment.

Talking point: "Natural language, no coding. AI coordinates the servers automatically."

Prompt 3: Show the Speed (20s)

How long would this analysis take manually vs. with MCP servers?

Key result: 40 hours -> 2-5 hours (production); 25-35 min in DRY_RUN demo. See Value Proposition for full metrics.

Prompt 4: Show the ROI (25s)

Calculate the annual ROI for analyzing 100 ovarian cancer patients per year,
assuming $6,000 traditional cost per patient.

Key result: ~$313K modeled annual savings at 100 patients; payback in first 2-3 patients.

Full PatientOne Demo (25-35 minutes)

End-to-end demo using all platform capabilities. For clinical details see PatientOne Profile.

Pathway 1: Quick Clinical Overview (5 min)

Using the mockepic server, summarize PAT001-OVC-2025's clinical history,
current medications, and treatment timeline.

Load PAT001-OVC-2025 genomic data and identify pathogenic variants
associated with ovarian cancer or treatment resistance.

Servers: mcp-mockepic, mcp-fgbio

Pathway 2: Multi-Omics Integration (15 min)

Load PAT001-OVC-2025 multi-omics data (RNA, protein, phospho) and run
Stouffer meta-analysis to identify consistently dysregulated genes.

Perform pathway enrichment on the top 50 hits. Focus on druggable pathways.

Servers: mcp-multiomics, mcp-fgbio

Pathway 3: Spatial Transcriptomics (15 min)

Load PAT001-OVC-2025 Visium spatial data. Run spatial differential expression
comparing tumor vs stromal regions, then pathway enrichment on spatially
variable genes.

Servers: mcp-spatialtools

Pathway 4: Full Platform E2E (45-60 min)

Uses all 19 servers including GEO, Open Targets, CIBERSORTx, neoantigen prediction. See Full E2E test prompt.

Post-Demo Talking Points

  • Safety: Every AI result requires clinician APPROVE/REVISE/REJECT
  • Status: Validated on synthetic data; clinical pilot is the next step
  • Timeline: 6 months to hospital pilot (assumes GCP + Azure AD in place)
  • Cost: $50K pilot (100 patients) or $75K/year production (500 patients)
  • Open source: Apache 2.0, no vendor lock-in on the server layer

See also: Executive Summary | Value Proposition | PatientOne Profile