Skip to main content
Challenge

Build an Evidence Review RAG Agent for Gene Function Publications

Extract gene identifiers and functional citations from research reports using NCBI gene packages. Search scientific summaries using Haystack to compile evidence matrices for biocuration teams.

Data ScienceHosted by Vera
Challenge brief

What you are building

The core problem, expected build, and operating context for this challenge.

Extract gene identifiers and functional citations from research reports using NCBI gene packages. Search scientific summaries using Haystack to compile evidence matrices for biocuration teams.

Scientific Evidence Reviewer

Ingest gene data reports and functional accession lists from NCBI Datasets packages.

Chunk and index accompanying literature abstracts in Haystack document stores.

Retrieve relevant experimental evidence for specific gene-protein pairings.

Synthesize an evidence summary table with provenance links for curator verification.

Multi-pane web interface showing source literature abstracts, matched accessions, and synthesized notes.

Use open-access PubMed Central articles matching NCBI gene IDs from public datasets download gene packages. Starter examples below are synthetic. They are not downloaded records.

NCBI Datasets gene download Relies on NCBI Datasets gene data report metadata containing gene IDs, symbols, and accession cross-references.

haystack, biopython, fastapi. Haystack provides modular retrieval-augmented generation pipelines tailored for document search and evidence synthesis. Hosted or local model supporting document QA via Haystack OpenAIGenerator or OllamaGenerator.

Submit runnable source, pinned dependencies, setup commands, a working interface, example outputs, and a recorded end-to-end demo. Include tests and an action log.

Extracts gene symbols and accessions from NCBI Datasets gene report JSON.

Haystack retrieval pipeline returns relevant abstracts with similarity scores.

Generates structured evidence summaries citing source document identifiers.

```json [ { "input": { "gene_symbol": "ACRV1", "literature_chunks": [ { "id": "art-1", "text": "ACRV1 encodes the acrosomal vesicle protein 1 expressed in human testis." } ] }, "expected_output": { "gene_symbol": "ACRV1", "matched_functions": [ "acrosomal vesicle protein 1 expression" ], "evidence_citations": [ "art-1" ], "curation_status": "CONFIRMED" } }, { "input": { "gene_symbol": "XYZ1", "literature_chunks": [] }, "expected_output": { "gene_symbol": "XYZ1", "matched_functions": [], "evidence_citations": [], "curation_status": "INSUFFICIENT_EVIDENCE" } } ] ```

Research literature curation tool with human-in-the-loop review; does not generate diagnostic advice. Use mock external writes. Require human review before any real deployment.

Rules

Participation guide

Research literature curation tool with human-in-the-loop review; does not generate diagnostic advice. Use mock external writes and human review.

Delivery guide

How work is evaluated

Evaluation

Extracts gene symbols and accessions from NCBI Datasets gene report JSON. Haystack retrieval pipeline returns relevant abstracts with similarity scores. Generates structured evidence summaries citing source document identifiers.

Submission format

Runnable source, pinned dependencies, setup commands, working interface, example outputs, recorded demo, tests, and action logs.

Datasets

Challenge datasets

Public datasets, plus the private datasets you have access to.

Evaluation rubric

How submissions are scored

These dimensions define what the evaluator checks and which criteria separate a passable run from a strong one.

Dimensions
4 scoring checks
Binary
1 pass or fail dimensions
Ordinal
3 scaled dimensions
Dimension 1

Complete workflow

Ingest gene data reports and functional accession lists from NCBI Datasets packages. Chunk and index accompanying literature abstracts in Haystack document stores. Retrieve relevant experimental evidence for specific gene-protein pairings. Synthesize an evidence summary table with provenance links for curator verification.

Score scale
012
Score 0

Missing or incorrect.

Score 1

Partly complete, with documented gaps.

Score 2

Complete, with reproducible evidence.

Dimension 2

Correctness and evidence

Extracts gene symbols and accessions from NCBI Datasets gene report JSON. Haystack retrieval pipeline returns relevant abstracts with similarity scores. Generates structured evidence summaries citing source document identifiers.

Score scale
012
Score 0

Missing or incorrect.

Score 1

Partly complete, with documented gaps.

Score 2

Complete, with reproducible evidence.

Dimension 3

Review interface

Multi-pane web interface showing source literature abstracts, matched accessions, and synthesized notes.

Score scale
012
Score 0

Missing or incorrect.

Score 1

Partly complete, with documented gaps.

Score 2

Complete, with reproducible evidence.

Dimension 4

Safe execution

Research literature curation tool with human-in-the-loop review; does not generate diagnostic advice.

Binary check

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

Learning goals

What you should walk away with

  • Ingest gene data reports and functional accession lists from NCBI Datasets packages.

  • Chunk and index accompanying literature abstracts in Haystack document stores.

  • Retrieve relevant experimental evidence for specific gene-protein pairings.

  • Synthesize an evidence summary table with provenance links for curator verification.

Resources and assets

Reference links and supporting material

Dataset notes

Use open-access PubMed Central articles matching NCBI gene IDs from public datasets download gene packages. Supplied starter examples are synthetic, not source extracts.

Challenge setup

Configuration summary for this challenge.

Challenge input

Input details are not configured.

Haystack

Framework

Evaluated output

No output checks are configured.

Evaluation checks
  • Ingest gene data reports and functional accession lists from NCBI Datasets packages. Chunk and index accomp...
  • Extracts gene symbols and accessions from NCBI Datasets gene report JSON. Haystack retrieval pipeline retur...
  • Multi-pane web interface showing source literature abstracts, matched accessions, and synthesized notes.
Configured evaluation evidence
  • 2 public reference cases
Configured runtime
  • 1 configured tool

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

Action Space
  • Haystack · Optional

Evaluation contract

  • Complete workflow · Weight 40
  • Correctness and evidence · Weight 30
  • Review interface · Weight 20
  • Safe execution · Weight 10

Reference cases

  • 2 public reference cases

Workflow tasks

No workflow tasks are configured.

Recipe freeze marker

No recipe freeze time is recorded.

Frequently Asked Questions about Build an Evidence Review RAG Agent for Gene Function Publications