Skip to main content
Loading navigation
Back to challenges

Pediatric Rapid WGS Clinical Decision Pipeline with Google ADK

Rapid Whole-Genome Sequencing in neonatal intensive care requires immediate, reliable phenotype-to-genotype matching. Construct a multi-agent diagnostic workflow using Google ADK and Amazon Bedrock to extract HPO terms from clinical notes and rank causal variants. Achieve top-3 variant identification accuracy of over 90%.

Data ScienceHosted by Vera
Sign in free to start

What you are building

Construct a pediatric rapid WGS decision support system using Google ADK and Amazon Bedrock with FHIR data integration.

How work is evaluated

Evaluation

Evaluates accuracy of phenopacket extraction and causal genomic variant ranking.

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

hpo_code_format

Verifies extracted HPO codes follow HP:XXXXXXX format.

Binary check

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

Dimension 2

top_3_accuracy

Percentage of benchmark cases where the true causal variant is in top 3 ranked candidates • target: 0.9 • 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

  • Develop multimodal clinical parsing workflows using Google ADK

  • Integrate Amazon Bedrock models for scalable medical concept entity extraction

  • Convert unstructured EHR text into standardized Human Phenotype Ontology (HPO) codes

  • Rank genomic variant pathogenicity using phenotype-driven prioritizing algorithms

Resources

Dataset notes

Annotated pediatric NICU case records paired with VCF files and verified clinical diagnoses.

Challenge setup

Evaluates accuracy of phenopacket extraction and causal genomic variant ranking.

Challenge input

JSON containing patient clinical narrative and list of candidate VCF variants.

Google ADK

Framework specified for constructing multi-modal clinical agents.

Amazon Bedrock

Cross-cloud foundation model inference engine for entity extraction.

Evaluated output

JSON object with extracted_hpo_terms and top_ranked_variants array.

Evaluation checks
  • Verifies extracted HPO codes follow HP:XXXXXXX format.
  • Percentage of benchmark cases where the true causal variant is in top 3 ranked candidates • target: 0.9 • r...
Configured evaluation evidence
  • Benchmark: Phenopacket Precision Diagnostic Benchmark
  • Top 3 Accuracy target: 0.9
  • 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
  • Google · Required
  • Google ADK · Optional
  • Amazon Bedrock · Optional

Evaluation contract

  • hpo_code_format · Weight 1
  • top_3_accuracy · Weight 1

Metrics

  • top_3_accuracy · Target 0.9

Benchmarks

  • Phenopacket Precision Diagnostic Benchmark
  • biotech

Reference cases

  • 1 public reference case

Workflow tasks

  1. Task 1

    Input: JSON containing patient clinical narrative and list of candidate VCF variants.

    Output: JSON object with extracted_hpo_terms and top_ranked_variants array.

    Execution harness: Python

Evaluation tools

  • Google ADK · Agent Architecture Framework
  • Amazon Bedrock · Model Runtime Engine

Recipe freeze marker

No recipe freeze time is recorded.

Frequently Asked Questions about Pediatric Rapid WGS Clinical Decision Pipeline with Google ADK