18 downstream tasks from the Nucleotide Transformer paper, providing a consistent genomics benchmark. The dataset, created by InstaDeepAI, is an updated version following peer review and was last updated on June 30, 2025.
Use Cases
- Benchmarking genomic sequence classification models based on the 18 provided tasks.
- Training binary classification models for specific genomic functions mentioned in the benchmark.
- Training multi-class classification models for genomic annotation tasks described in the paper.
- Evaluating the transfer learning capabilities of foundation models on genomics data.
Strengths
- Provides a consistent benchmark of 18 tasks for genomics model evaluation.
- Is an updated version following peer review of the original Nucleotide Transformer paper.
- Created by InstaDeepAI, a known organization in AI for science.
Limitations
- Column-level documentation is absent; field semantics must be inferred after download.
- Row count is unknown, which may limit suitability assessment.
- Description metadata is limited; actual data quality requires manual inspection after download.
Provenance
- Source
- InstaDeepAI
- Collection Method
- Tasks presented in the Nucleotide Transformer paper.
- Time Range
- null
- Freshness
- Last updated 2025-06-30 10:37:24; freshness should be verified.
- Geography
- null