Nanopore Long-Read Structural Variant Indexer with LlamaIndex and AI21 Studio
Oxford Nanopore long-read sequencing unlocks structural variants across biopharma workflows. You will build a RAG pipeline using LlamaIndex and AI21 Studio to query structural variant databases and genomic documentation, achieving an explicit query retrieval precision of over 88%.
What you are building
Build a genomic RAG system using LlamaIndex and AI21 Studio to index and query Nanopore long-read structural variants.
How work is evaluated
Evaluates genomic RAG retrieval precision and answer factual precision on Nanopore SV queries.
Datasets
Public datasets, plus the private datasets you have access to.
How submissions are scored
variant_id_match
Ensure correct structural variant ID is returned
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
retrieval_precision
Precision of relevant genomic variants retrieved • target: 0.88 • range: 0-1
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
Learning goals
Construct knowledge indices over VCF genomic files using LlamaIndex framework
Integrate AI21 Studio language models for specialized biological information retrieval
Execute natural language queries for structural variant functional impact identification
Resources
Oxford Nanopore genomic structural variant VCF dataset with 1,000 annotated biopharma clinical samples.
Challenge setup
Evaluates genomic RAG retrieval precision and answer factual precision on Nanopore SV queries.
Challenge input
JSON object with query_string
LlamaIndex
Mandatory framework for biological data retrieval.
AI21 Studio
Provides language model reasoning capabilities.
Evaluated output
JSON object with answer string and retrieved_variant_ids list
- Ensure correct structural variant ID is returned
- Precision of relevant genomic variants retrieved • target: 0.88 • range: 0-1
- Benchmark: GenoRAG-Bench
- Retrieval Precision target: 0.88
- 1 public reference case
- Python execution harness
- Python sandbox (unavailable on Versalist)
Shows the first task, up to three tools, and up to three evidence items.
View technical recipe
Configured tools
- AI21 Studio · Required
- Llama Index · Optional
- Exa · Optional
Evaluation contract
- variant_id_match · Weight 1
- retrieval_precision · Weight 1
Metrics
- retrieval_precision · Target 0.88
Benchmarks
- GenoRAG-Bench
- biotech
Reference cases
- 1 public reference case
Workflow tasks
- Task 1
Input: JSON object with query_string
Output: JSON object with answer string and retrieved_variant_ids list
Execution harness: Python
Evaluation tools
- LlamaIndex · RAG Framework
- AI21 Studio · Language Model Endpoint
Recipe freeze marker
No recipe freeze time is recorded.