Skip to main content
Loading navigation
Back to challenges

Clinical Cancer Mutant Dosage Stratification Agent with Claude Agents SDK

Calculating gene mutant dosage across large oncology cohorts is critical for predicting metastatic disease progression. Build a Claude Agents SDK workflow utilizing Llama 3.3 70B to parse clinical cancer genomic profiles and compute mutant dosage risk scores for target organ tropism.

Data ScienceHosted by Vera
Sign in free to start

What you are building

Implement an automated oncology analytics workflow using Claude Agents SDK and Llama 3.3 70B to evaluate mutant allele dosage across 60,000 tumor profiles.

How work is evaluated

Evaluation

Evaluates mathematical accuracy of calculated mutant dosage scores and predicted metastatic organ risk.

Datasets

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

How submissions are scored

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

Dosage Score Within Tolerances

Ensures calculated dosage equals vaf * copy_number.

Binary check

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

Dimension 2

Tropism Prediction AUC

Area under curve for predicting organ metastatic tropism. • target: 0.85 • range: 0.5-1

Binary check

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

Learning goals

  • Construct extended agentic workflows using Claude Agents SDK.

  • Leverage Llama 3.3 70B inference for quantitative allele dosage parsing.

  • Correlate copy-number changes and somatic mutations with metastatic organ tropism.

  • Export structured oncological stratification matrices.

Resources

Dataset notes

Oncology patient cohort subset featuring VAF, copy number annotations, primary site descriptions, and verified metastatic outcomes.

Challenge setup

Evaluates mathematical accuracy of calculated mutant dosage scores and predicted metastatic organ risk.

Challenge input

JSON containing sample ID, gene, variant allele frequency (VAF), copy number, and tissue origin

Claude Agents SDK

Provides extended thinking framework for analytical workflows.

Llama 3.3 70B

High-capacity open model for quantitative parsing.

Evaluated output

JSON containing computed mutant dosage, risk category, and predicted metastatic site

Evaluation checks
  • Ensures calculated dosage equals vaf * copy_number.
  • Area under curve for predicting organ metastatic tropism. • target: 0.85 • range: 0.5-1
Configured evaluation evidence
  • Benchmark: OncoGenomics-Bench
  • Tropism Prediction AUC target: 0.85
  • 1 public reference case
Configured runtime
  • Python execution harness
  • Python sandbox (unavailable on Versalist)

Shows the first task, up to three tools, and up to three evidence items.

View technical recipe

Configured tools

Action Space
  • Llama 3.3 70B · Required
  • Meta · Optional
  • Featuretools · Optional

Evaluation contract

  • Dosage Score Within Tolerances · Weight 1
  • Tropism Prediction AUC · Weight 1

Metrics

  • Tropism Prediction AUC · Target 0.85

Benchmarks

  • OncoGenomics-Bench
  • biotech

Reference cases

  • 1 public reference case

Workflow tasks

  1. Task 1

    Input: JSON containing sample ID, gene, variant allele frequency (VAF), copy number, and tissue origin

    Output: JSON containing computed mutant dosage, risk category, and predicted metastatic site

    Execution harness: Python

Evaluation tools

  • Claude Agents SDK · Agent Orchestration
  • Llama 3.3 70B · Inference Model

Recipe freeze marker

No recipe freeze time is recorded.

Frequently Asked Questions about Clinical Cancer Mutant Dosage Stratification Agent with Claude Agents SDK