Pediatric Rapid WGS Decision Agent using Google ADK and o3
Rapid whole-genome sequencing in pediatric ICU settings demands immediate, high-confidence clinical phenotype mapping. Developers will build a Google ADK agent powered by o3 to cross-reference patient HPO terms with pediatric rWGS variant files, generating rapid diagnostic candidate rankings within a 30-second processing window.
What you are building
The core problem, expected build, and operating context for this challenge.
Develop an automated pediatric rWGS diagnostic tool using Google ADK and o3 reasoning model to match HPO terms with genomic variants.
How work is evaluated
Evaluates candidate gene ranking recall, HPO match accuracy, and rapid processing execution speed.
Challenge datasets
Public datasets, plus the private datasets you have access to.
How submissions are scored
These dimensions define what the evaluator checks and which criteria separate a passable run from a strong one.
top_gene_correctness
Validates that SCN1A is identified as the primary candidate gene.
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
top_1_recall
Recall of true causal gene at position 1. • target: 0.95 • range: 0-1
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
What you should walk away with
Configure Google Agent Development Kit (ADK) agent with o3 model reasoning.
Parse Human Phenotype Ontology (HPO) terms from clinical ICU unstructured notes.
Cross-match HPO codes against whole-genome variant call format (VCF) data.
Generate prioritized candidate gene rankings for rare pediatric diseases.
Reference links and supporting material
Pediatric clinical vignettes with associated anonymized rWGS variant files and verified diagnoses.
Challenge setup
Evaluates candidate gene ranking recall, HPO match accuracy, and rapid processing execution speed.
Challenge input
JSON containing clinical HPO terms and candidate VCF variant list.
Google ADK
Provides robust structure for building complex multi-step agents.
o3
Delivers advanced multi-step logic needed for complex phenotype-genotype mapping.
Evaluated output
JSON containing top 3 ranked candidate genes and matched phenotype justifications.
- Validates that SCN1A is identified as the primary candidate gene.
- Recall of true causal gene at position 1. • target: 0.95 • range: 0-1
- Benchmark: Rare Disease Diagnostics Benchmark
- Top 1 Recall target: 0.95
- 1 public reference case
- 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
- Google · Optional
- Google ADK · Optional
- o3 · Required
Evaluation contract
- top_gene_correctness · Weight 1
- top_1_recall · Weight 1
Metrics
- top_1_recall · Target 0.95
Benchmarks
- Rare Disease Diagnostics Benchmark
- biotech
Reference cases
- 1 public reference case
Workflow tasks
- Task 1
Input: JSON containing clinical HPO terms and candidate VCF variant list.
Output: JSON containing top 3 ranked candidate genes and matched phenotype justifications.
Execution harness: Python
Evaluation tools
- Google ADK · Agent Development Framework
- o3 · Reasoning Language Model
Recipe freeze marker
No recipe freeze time is recorded.