00panel — Proteomic Panel Design

Overview

00panel is a decision-support system for designing new Olink proteomic panels. It answers the question: given thousands of measurable proteins, which combination should go into a new panel?

The system integrates demand signals (who wants it measured?), technical feasibility (can we build it?), genetic evidence (is it causally important?), and competitive positioning (does it differentiate us?) into a multi-criteria optimization framework.

graph TB
    subgraph Input["Input"]
        SEEDS["Disease category<br/>or seed proteins"]
    end

    subgraph Expand["Discovery"]
        direction TB
        REACT["Reactome pathway<br/>co-membership"]
        KG["Knowledge Graph<br/>multi-hop traversal"]
    end

    subgraph Score["Scoring (4 dimensions, normalized 0–1)"]
        direction TB
        D["Demand<br/>NIH grants · Pub growth<br/>Clinical trials · GWAS"]
        E["Evidence<br/>pQTL colocalisation<br/>KG disease links"]
        F["Feasibility<br/>Blood-detectable<br/>Concentration range"]
        U["Uniqueness<br/>Competitor coverage<br/>Differentiation"]
    end

    subgraph Optimize["Optimization"]
        OPT["Greedy selection<br/>+ diversity constraints<br/>+ Pareto front"]
    end

    subgraph Validate["Validation"]
        BT["Backtest vs<br/>publication history"]
        CMP["Compare vs<br/>5 competitors"]
    end

    subgraph Output["Output"]
        PANEL["Proposed Panel<br/>ranked proteins + UniProt IDs<br/>+ statistical justification"]
    end

    Input --> Expand
    Expand --> Score
    D & E & F & U --> OPT
    OPT --> Validate
    Validate --> Output

How it works

sequenceDiagram
    participant U as User
    participant T as TUI / CLI
    participant S as Signal Collectors
    participant API as External APIs
    participant D as Data Layer
    participant O as Optimizer

    U->>T: Select disease or enter genes
    T->>D: Expand via Reactome pathways
    D-->>T: 50+ candidates
    T->>S: Score all candidates
    S->>API: NIH · OpenAlex · CT.gov · GWAS (parallel)
    API-->>S: Raw counts
    S-->>T: Normalized scores [0-1]
    T->>D: Check feasibility + competitors
    D-->>T: Blood-detectable? Unique?
    T->>O: Multi-criteria optimize
    O-->>T: Top N proteins
    T-->>U: Panel + stats + validation

Evidence Tiers

The system organizes analyses by trustworthiness:

Tier	What it does	Reliability
OBSERVE	Raw data lookups (grants, publications, trials)	Factual, independently verifiable
COMPARE	Set operations (competitor diff, pathway coverage %)	Deterministic, no inference
PROPOSE	Multi-criteria optimization with modeling choices	Depends on weight configuration
DEEP	KG multi-hop inference, structural similarity	Requires graph data, highest insight potential

Quick example

$ panel demo alzheimer --size 5

  Disease: alzheimer — 10 seed proteins
  Expanded to 50 candidates via pathways
  Scored across 4 dimensions (20 API calls)

  Proposed 5-protein panel:
    APP  (P05067)  — demand=1.000, highest NIH funding
    APOE (P02649)  — demand=0.746, +40% pub growth
    GFAP (P14136)  — demand=0.322, rising biomarker
    TREM2 (Q9NZC2) — demand=0.160, genetic evidence
    MAPT (P10636)  — demand=0.105, established target

System Architecture

graph LR
    subgraph 00panel
        CLI[CLI / TUI]
        SIG[signals/]
        FEAS[feasibility/]
        SEL[selection/]
        DATA[data/]
        EV[evidence/]
    end

    subgraph External["External APIs (live)"]
        NIH[NIH RePORTER]
        OA[OpenAlex]
        CT[ClinicalTrials.gov]
        GWAS[GWAS Catalog]
        OT[Open Targets Genetics]
    end

    subgraph bioingest["bioingest (local data)"]
        COMP[Competitor menus]
        PATH[Reactome pathways]
        MAP[Protein mapping 26K]
        HPA[HPA secretome]
    end

    subgraph graphrag["graphrag_api (KG)"]
        NEP[Neptune<br/>70K nodes · 263K rels]
    end

    SIG --> NIH & OA & CT & GWAS
    EV --> OT & NEP
    DATA --> COMP & PATH & MAP & HPA

Launch modes

Command	Mode
uv run 00panel	Interactive TUI (arrow-key navigation)
panel tui	Same TUI via CLI
panel demo alzheimer	Quick disease-based demo
panel score TNF IL6	Direct CLI command
panel --help	All available commands

Data sources

Source	Signal	Access
NIH RePORTER	Active grants (demand)	REST API, live
OpenAlex	Publication YoY growth	REST API, live
ClinicalTrials.gov	Phase 2+ trials	REST API, live
GWAS Catalog	SNP-trait associations	REST API, live
Open Targets Genetics	pQTL colocalisation	GraphQL, live
HPA Secretome	Blood detectability	Cached download
Reactome	Pathway membership	bioingest bulk
Competitor menus	5 platforms	bioingest bulk
Knowledge Graph	40+ sources integrated	graphrag_api