Automated Western Blot Integrity Auditor with Mastra AI and OpenAI o3
Emerging reports highlight widespread image manipulation in scientific antibody validation datasets. Build a multi-modal audit pipeline with Mastra AI and OpenAI o3 to inspect biomedical image artifacts and flag duplications or spliced gel bands. Achieve a 0.95 F1-score across benchmark scientific literature image sets.
What you are building
The core problem, expected build, and operating context for this challenge.
Implement a TypeScript-based scientific image audit system using Mastra AI and OpenAI o3 for detecting antibody validation manipulation.
How work is evaluated
Evaluates image manipulation detection accuracy on Western blot scientific figure benchmarks.
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.
schema_validity
Output contains manipulation_detected boolean and confidence score.
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
f1_score
F1 metric for identifying altered scientific images • 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
Build multi-modal image analysis workflows using Mastra AI framework
Leverage OpenAI o3 reasoning capabilities for detecting subtle pixel duplications and band splices
Generate automated forensic audit reports for paper retraction workflows
Integrate OpenCV canvas processing in Node.js/TypeScript environments
Reference links and supporting material
Dataset of 200 synthetic and annotated Western blot gel images containing authentic controls and manipulated duplicates/splices.
Challenge setup
Evaluates image manipulation detection accuracy on Western blot scientific figure benchmarks.
Challenge input
JSON containing image URLs or base64 encoded gel images.
Mastra AI
Required framework for building structured agent workflows in TypeScript.
OpenAI o3
Advanced visual reasoning model for pixel-level anomaly verification.
Evaluated output
JSON object with manipulation_detected boolean, anomaly_locations, and reasoning trace.
- Output contains manipulation_detected boolean and confidence score.
- F1 metric for identifying altered scientific images • target: 0.95 • range: 0-1
- Benchmark: Scientific Figure Image Integrity Benchmark
- F1 Score target: 0.95
- 1 public reference case
- JavaScript execution harness
- javascript 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
- OpenAI o3 · Required
- OpenAI · Optional
- Mastra AI · Optional
- Mastra AI · Optional
Evaluation contract
- schema_validity · Weight 1
- f1_score · Weight 1
Metrics
- f1_score · Target 0.95
Benchmarks
- Scientific Figure Image Integrity Benchmark
- biotech
Reference cases
- 1 public reference case
Workflow tasks
- Task 1
Input: JSON containing image URLs or base64 encoded gel images.
Output: JSON object with manipulation_detected boolean, anomaly_locations, and reasoning trace.
Execution harness: JavaScript
Evaluation tools
- Mastra AI · TypeScript Agent Framework
- OpenAI o3 · Reasoning Multimodal Model
Recipe freeze marker
No recipe freeze time is recorded.