agentleFS
aws-samplesaws-samples/amazon-bedrock-agents-healthcare-lifesciences/skills/biomarker-multi-agent-discovery/SKILL.md

biomarker-multi-agent-discovery

Skill277 starsChanged 6 months agoMIT-0View on GitHub

Use when orchestrating a multi-agent biomarker discovery workflow that requires coordinating database queries, pathway analysis, literature review, statistical modeling, and clinical evidence synthesis to produce ranked biomarker panels.

The file

We split this skill into the rules it gives an agent

3 rules. About 0 are common, 2 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Call biomarker_database_analyst_agent with the data retrieval question

    line 48
  2. Distinctive:

    Always explain the multi-step plan to the user before executing

    line 127
  3. Distinctive:

    When agents fail, explain which step failed and what alternatives exist

    line 130

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Biomarker Multi-Agent Discovery

When to use this skill

  • Complex biomarker discovery requiring multiple analysis modalities
  • Coordinating database queries with statistical survival analysis
  • Synthesizing findings from literature, pathways, and clinical data into a biomarker panel
  • Questions that span multiple sub-domains (e.g., "find best biomarker for survival in chemo patients and show evidence")

Architecture: Agents-as-Tools Pattern

The orchestrator dispatches to specialized sub-agents, each wrapped as a tool:

Orchestrator (Supervisor)
  |-- biomarker_database_analyst_agent  -> SQL queries on clinical genomic data
  |-- clinical_evidence_research_agent  -> PubMed + Knowledge Base search
  |-- statistician_agent                -> Survival regression, Kaplan-Meier plots
  |-- medical_imaging_agent             -> Radiomics biomarker extraction

Cross-agent data sharing uses AgentCore Memory: the database agent stores query results, and downstream agents (statistician) retrieve them automatically.

Orchestration Workflow

Step 1: Classify the user query

Map the question to required sub-agents:

Query typeAgents neededSequence
Demographics / countsDatabase analyst onlySingle call
Literature evidenceClinical evidence researcher onlySingle call
Statistical analysis (p-values, survival)Database analyst -> StatisticianSequential
Imaging biomarkersDatabase analyst -> Medical imagingSequential
Comprehensive discoveryAll agentsMulti-step
Pathway interpretationDatabase analyst -> LiteratureSequential

Step 2: Execute database queries first

For any analysis requiring patient data:

  1. Call biomarker_database_analyst_agent with the data retrieval question
  2. Results are automatically stored in shared memory
  3. Include required columns: survival_status, survival_duration, biomarker expression values

Example dispatch:

Query: "What are the top 5 biomarkers with overall survival for chemo patients?"
-> Database agent: "Query all records including survival status, survival duration in years, and gene expression values for patients where chemotherapy = 'Yes'"

Step 3: Feed results to downstream agents

For statistical analysis:

-> Statistician agent: "Fit a survival regression model on the query results"

The statistician retrieves data from memory automatically. No S3 path needed.

For visualization:

-> Statistician agent: "Generate a bar chart of the top 5 biomarkers by p-value"
-> Statistician agent: "Plot Kaplan-Meier curve for GDF15 with threshold 10"

For literature validation:

-> Clinical evidence researcher: "Search PubMed for evidence on GDF15 as a biomarker in NSCLC"

Step 4: Synthesize findings into ranked biomarker panel

Combine outputs from all agents into a consolidated report:

Biomarker Panel Report
=====================
1. [Gene] - p-value: X, HR: Y
   - Pathway: [from pathway analysis]
   - Literature: [N publications supporting]
   - Clinical significance: [interpretation]

2. [Gene] - p-value: X, HR: Y
   ...

Ranking criteria (in priority order):

  1. Statistical significance (lowest p-value from Cox regression)
  2. Clinical significance (hazard ratio magnitude)
  3. Pathway relevance (membership in disease-associated pathways)
  4. Literature support (number of supporting publications)
  5. Biological plausibility (protein function matches disease mechanism)

Step 5: Generate actionable recommendations

For each top biomarker, provide:

  • Measurement method (IHC, RNA-seq, blood test)
  • Patient stratification threshold (expression cutoff)
  • Potential clinical utility (prognostic vs. predictive vs. diagnostic)
  • Next validation steps (cohort size, assay development)

Tool Dispatch Reference

AgentToolsInputOutput
Database Analystget_schema, query_redshift, refine_sqlData questionQuery results (auto-stored in memory)
Clinical Evidencequery_pubmed, retrieveEvidence questionLiterature summary with citations
Statisticianrun_code, plot_kaplan_meier, fit_survival_regressionAnalysis requestRegression, charts (S3 paths), p-values
Medical Imagingcompute_imaging_biomarker, analyze_imaging_biomarkerPatient IDsRadiomics features (sphericity, elongation)

Example Multi-Step Sequences

"Find best biomarker for survival in chemo patients with visualization":

  1. Database agent -> Statistician (Cox regression) -> Statistician (Kaplan-Meier) -> Literature -> Synthesize

"Compare imaging biomarkers for patients with lowest GDF15":

  1. Database agent (find patients) -> Medical imaging (compute + visualize) -> Synthesize

Conventions

  • Always explain the multi-step plan to the user before executing
  • Present results from each agent separately, then provide consolidated summary
  • Include S3 paths for any generated charts or images
  • When agents fail, explain which step failed and what alternatives exist
  • Medical/statistical concepts must be explained in accessible language
  • Memory events expire after 3 days -- no manual cleanup needed

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.