Precision Target Prioritization Agent
Design a target-disease prioritization system that leverages the Claude Agents SDK's 'Extended Thinking' capabilities to reason over complex genetic evidence. Your agent will interface with the Open Targets GraphQL API to retrieve genetic constraint scores (pLI), target-disease association scores, and drug tractability data. Using the o4-mini model for rapid data parsing within the agent loop, the system must prioritize therapeutic targets for specific diseases based on evidence strength and clinical tractability. The challenge involves writing a GraphQL planner that dynamically constructs queries to minimize data transfer while maximizing information gain (e.g., fetching L2G scores for specific variants). Claude will then synthesize this data to explain 'why' a specific target is prioritized, citing genetic scores and clinical trial status.
What you are building
The core problem, expected build, and operating context for this challenge.
Build a GraphQL-aware agent with Claude Agents SDK to prioritize therapeutic targets by integrating Open Targets genetic and tractability data.
How work is evaluated
Evaluation will focus on the agent's ability to rank a list of targets correctly according to a specific scoring weighted formula provided in the prompt.
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.
GraphQL Validity
Ensures the agent generates syntactically correct GraphQL queries.
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
Ranking Precision
Correlation between agent ranking and Open Targets platform ranking. • target: 0.9 • 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
Build a GraphQL-native tool-calling agent using Claude Agents SDK
Implement 'Extended Thinking' blocks to weigh conflicting genetic evidence from Open Targets
Optimize data retrieval using a GraphQL planner to select only high-impact fields (e.g., score.overall)
Analyze drug tractability and modality (small molecule vs antibody) using structured Open Targets schemas
Reference links and supporting material
Sample data for 1 tasks
Challenge setup
Evaluation will focus on the agent's ability to rank a list of targets correctly according to a specific scoring weighted formula provided in the prompt.
Challenge input
Disease name: 'Alzheimer's Disease'
Zed
High-performance code editor
RAI
Agentic framework for robotics using ROS 2
o4-mini
Policy Serving in the agent workflow.
Evaluated output
Ordered JSON list of target symbols with justification
- Ensures the agent generates syntactically correct GraphQL queries.
- Correlation between agent ranking and Open Targets platform ranking. • target: 0.9 • range: 0-1
- Ranking Precision target: 0.9
- 1 public reference case
- Python execution harness
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
- Zed · Optional
- RAI · Optional
- o4-mini · Required
Evaluation contract
- GraphQL Validity · Weight 1
- Ranking Precision · Weight 1
Metrics
- Ranking Precision · Target 0.9
Reference cases
- 1 public reference case
Workflow tasks
- Task 1
Input: Disease name: 'Alzheimer's Disease'
Output: Ordered JSON list of target symbols with justification
Execution harness: Python
Recipe freeze marker
No recipe freeze time is recorded.