Build a Variant Functional Effect Predictor with OpenAI Agents SDK
Missense mutation interpretation requires fast, accurate functional predictions combined with automated triage. Developers will build an OpenAI Agents SDK pipeline integrated with Bolt.new web workflows to ingest genomic VCF files, query missense prediction endpoints, and output structured ACMG classification reports with 95% evaluation accuracy.
What you are building
The core problem, expected build, and operating context for this challenge.
Implement an OpenAI Agents SDK workflow that orchestrates genomic variant annotations and evaluates missense mutations.
How work is evaluated
Evaluates variant triage accuracy, ACMG code assignment precision, and JSON output formatting.
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.
json_schema_validation
Validates that the output matches the required ACMG report structure.
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
classification_accuracy
Accuracy of pathogenic vs benign classification compared to ground truth. • 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
Ingest VCF variant files and extract missense mutation annotations using Biopython.
Build tool-calling agents with OpenAI Agents SDK to query functional prediction tools.
Synthesize genomic evidence into ACMG-compliant structured JSON reports.
Integrate agent runtime endpoints with Bolt.new frontends for interactive variant triage.
Reference links and supporting material
Synthetic and open-source genomic variant datasets containing ClinVar annotated missense mutations.
Challenge setup
Evaluates variant triage accuracy, ACMG code assignment precision, and JSON output formatting.
Challenge input
JSON containing genomic variant coordinates and reference/alternate alleles.
OpenAI Agents SDK
Official framework for multi-turn tool-calling agents.
Bolt.new
AI engineering tool for rapidly deploying reactive web interfaces.
Evaluated output
JSON containing ACMG classification, pathogenicity score, and evidence codes.
- Validates that the output matches the required ACMG report structure.
- Accuracy of pathogenic vs benign classification compared to ground truth. • target: 0.95 • range: 0-1
- Benchmark: ClinVar Variant Benchmarks
- Classification Accuracy 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
- Bolt.new · Optional
- OpenAI · Required
- Bolt.new · Optional
- Edge Impulse · Optional
Evaluation contract
- json_schema_validation · Weight 1
- classification_accuracy · Weight 1
Metrics
- classification_accuracy · Target 0.95
Benchmarks
- ClinVar Variant Benchmarks
- biotech
Reference cases
- 1 public reference case
Workflow tasks
- Task 1
Input: JSON containing genomic variant coordinates and reference/alternate alleles.
Output: JSON containing ACMG classification, pathogenicity score, and evidence codes.
Execution harness: Python
Evaluation tools
- OpenAI Agents SDK · Agent Orchestration Framework
- Bolt.new · Frontend Prototyping
Recipe freeze marker
No recipe freeze time is recorded.