Skip to main content
Back to challenges
Challenge

Design a Leptin Biomarker Trial Agent with Claude Agents SDK & Claude 4.1 Opus

Evaluating exercise-mimetic drugs requires precise longitudinal extraction of leptin sensitivity and muscle mass indicators. Build an analytical extraction agent using the Claude Agents SDK and Claude 4.1 Opus to automatically summarize clinical trial logs and identify efficacy metrics with 95% classification precision.

Data ScienceHosted by Vera
Sign in free to start
Challenge brief

What you are building

The core problem, expected build, and operating context for this challenge.

Construct an analytical trial processing pipeline leveraging Claude Agents SDK and Claude 4.1 Opus for clinical biomarker extraction.

Delivery guide

How work is evaluated

Evaluation

Evaluates biomarker extraction precision and recall across Phase 1 patient trial reports.

Datasets

Challenge datasets

Public datasets, plus the private datasets you have access to.

Evaluation rubric

How submissions are scored

These dimensions define what the evaluator checks and which criteria separate a passable run from a strong one.

Dimensions
2 scoring checks
Binary
2 pass or fail dimensions
Ordinal
0 scaled dimensions
Dimension 1

exact_extraction_accuracy

Verifies extracted figures match within 0.1 tolerance.

Binary check

This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.

Dimension 2

extraction_f1_score

Precision and recall of extracted biomedical entities. • target: 0.95 • range: 0-1

Binary check

This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.

Learning goals

What you should walk away with

  • Deploy Claude Agents SDK with extended thinking enabled.

  • Extract structured muscle mass and leptin biomarker metrics from unstructured trial notes.

  • Verify extraction fidelity using gold-standard clinical trial benchmark datasets.

Resources and assets

Reference links and supporting material

Dataset notes

Dataset of 50 synthetic Phase 1 clinical trial progress reports for metabolic and exercise-mimetic compounds.

Challenge setup

Evaluates biomarker extraction precision and recall across Phase 1 patient trial reports.

Challenge input

Unstructured text document

Claude Agents SDK

Primary agent building framework specified in requirements.

Claude 4.1 Opus

Provides reasoning for complex clinical document extraction.

Evaluated output

JSON containing lean_mass_change_pct and leptin_sensitivity_index

Evaluation checks
  • Verifies extracted figures match within 0.1 tolerance.
  • Precision and recall of extracted biomedical entities. • target: 0.95 • range: 0-1
Configured evaluation evidence
  • Benchmark: BLURB-ClinicalExtraction
  • Extraction F1 Score target: 0.95
  • 1 public reference case
Configured runtime
  • Python execution harness
  • Python sandbox (unavailable on Versalist)

Summary shows the first task, up to three tools, and up to three evidence items. Open the technical recipe to view all configured items.

View technical recipe

Configured tools

Action Space
  • Ogre.run · Optional
  • Featherless AI · Optional
Policy Serving
  • Claude 4.1 Opus · Required

Evaluation contract

  • exact_extraction_accuracy · Weight 1
  • extraction_f1_score · Weight 1

Metrics

  • extraction_f1_score · Target 0.95

Benchmarks

  • BLURB-ClinicalExtraction
  • biotech

Reference cases

  • 1 public reference case

Workflow tasks

  1. Task 1

    Input: Unstructured text document

    Output: JSON containing lean_mass_change_pct and leptin_sensitivity_index

    Execution harness: Python

Evaluation tools

  • Claude Agents SDK · agent_framework
  • Claude 4.1 Opus · llm

Recipe freeze marker

No recipe freeze time is recorded.

Frequently Asked Questions about Design a Leptin Biomarker Trial Agent with Claude Agents SDK & Claude 4.1 Opus