Nucleotide Transformer: building and evaluating robust foundation models for human genomics (Nature Methods)
Nature Methods 22, 287-297 (published online 2024-11-28); PMC11810778 full-text XML · Peer-reviewed
What this source contributed
Records in this release that cite this source. A record can cite several sources.
Results
Values as the source printed them. Compare results only under compatible protocols, datasets and metrics.
| Value | Metric | Model or configuration | Dataset | Evaluation |
|---|---|---|---|---|
| 50 minute | Runtime (Mean wall time per IA3 fine-tuning run) | Nucleotide Transformer 2.5B-parameter models, IA3 fine-tuning (Dalla-Torre et al. 2025) | NT benchmark, all 18 tasks | NT 2.5B-parameter models IA3 fine-tuning run time (Dalla-Torre et al. 2025) |
| 20 minute | Runtime (Mean wall time per IA3 fine-tuning run) | Nucleotide Transformer 500M-parameter models, IA3 fine-tuning (Dalla-Torre et al. 2025) | NT benchmark, all 18 tasks | NT 500M-parameter models IA3 fine-tuning run time (Dalla-Torre et al. 2025) |
Claims
Statements this source supports, each with where in the source it was checked.
- Nucleotide Transformer fine-tuning time on a single GPU (Discussion): Fine tuning time The Discussion states that IA3 fine-tuning, which trains about 0.1% of parameters, allows even the largest models to be fine-tuned in under 15 min on a single GPU. This conflicts with Methods 'Hardware' (one node of eight A100 GPUs, 20 min average for 500M and 50 min for 2.5B models). (Discussion, paragraph 3: 'allowing even our largest models to be fine-tuned in under 15 min on a single GPU.')
- Nucleotide Transformer pre-training and probing compute: Training compute All models were trained on Cambridge-1 (16 nodes of eight A100 GPUs, 128 GPUs). The 500M-parameter models trained on one node for a day; the 2.5B-parameter models used the whole cluster for 28 days; NT-v2 models (50M to 500M) trained on one node for a day. Probing embeddings took 2 days on one 8-A100 node, and fitting 760,000 downstream models took 2.5 days on 3,000 CPUs. (Methods, 'Data preparation > Hardware', paragraphs 1 and 2: 'The 500M-parameter models were trained on a single node for a day, whereas the 2.5B-parameter models required the whole cluster for 28 days to be trained.')
- Relevance of Nucleotide Transformer IA3 fine-tuning run time (Dalla-Torre et al. 2025 Methods 'Hardware') to "Select an execution workflow for large-scale diagnostic genomics": Assessed by Nucleotide Transformer IA3 fine-tuning run time (Dalla-Torre et al. 2025 Methods 'Hardware') (gfm-guide-20261014-source-nt-nature-methods: Methods 'Data preparation > Hardware', paragraph 1; Discussion, paragraph 3)
Models, benchmarks and other records
Configurations (2)
Protocols (1)
Evaluations (2)
Evidence
Source checking verifies the cited claim or transcription. It does not establish independent reproduction.
Evidence table
Inspect claims, sources and review details
Trace each statement to its source and review. A context-only reference supports the record generally; it does not verify an individual field. Source checking does not reproduce an experiment.
One row per statement and cited source. Multiple citations are not independent evaluations. Shared locators are labelled explicitly.
18 evidence rows matching the loaded filters
| Property and statement | Original source and location | Review and provenance |
|---|---|---|
| attributes.artifact_sha256 7c3b78a4f38ef053a08e466222e5662dd9711d91c535512d0b35d459d1fc7249 Source metadata | Nucleotide Transformer: building and evaluating robust foundation models for human genomics (Nature Methods) No field-specific location recorded Version: Nature Methods 22, 287-297 (published online 2024-11-28); PMC11810778 full-text XML | catalogued No individual claim review recorded Source has a recorded evidence concern. Consult its source page before using the claim. Audit detailsField: Source artifact SHA-256: Hash scope: Hash scope not separately documented; inspect source record |
| Artifact url https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11810778/fullTextXML Source metadata | Nucleotide Transformer: building and evaluating robust foundation models for human genomics (Nature Methods) No field-specific location recorded Version: Nature Methods 22, 287-297 (published online 2024-11-28); PMC11810778 full-text XML | catalogued No individual claim review recorded Source has a recorded evidence concern. Consult its source page before using the claim. Audit detailsField: Source artifact SHA-256: Hash scope: Hash scope not separately documented; inspect source record |
| Doi 10.1038/s41592-024-02523-z Source metadata | Nucleotide Transformer: building and evaluating robust foundation models for human genomics (Nature Methods) No field-specific location recorded Version: Nature Methods 22, 287-297 (published online 2024-11-28); PMC11810778 full-text XML | catalogued No individual claim review recorded Source has a recorded evidence concern. Consult its source page before using the claim. Audit detailsField: Source artifact SHA-256: Hash scope: Hash scope not separately documented; inspect source record |
Evidence concerns1 values
| Nucleotide Transformer: building and evaluating robust foundation models for human genomics (Nature Methods) No field-specific location recorded Version: Nature Methods 22, 287-297 (published online 2024-11-28); PMC11810778 full-text XML | catalogued No individual claim review recorded Source has a recorded evidence concern. Consult its source page before using the claim. Audit detailsField: Source artifact SHA-256: Hash scope: Hash scope not separately documented; inspect source record |
| Licence CC-BY-NC-ND-4.0 Source metadata | Nucleotide Transformer: building and evaluating robust foundation models for human genomics (Nature Methods) No field-specific location recorded Version: Nature Methods 22, 287-297 (published online 2024-11-28); PMC11810778 full-text XML | catalogued No individual claim review recorded Source has a recorded evidence concern. Consult its source page before using the claim. Audit detailsField: Source artifact SHA-256: Hash scope: Hash scope not separately documented; inspect source record |
| Media type application/xml Source metadata | Nucleotide Transformer: building and evaluating robust foundation models for human genomics (Nature Methods) No field-specific location recorded Version: Nature Methods 22, 287-297 (published online 2024-11-28); PMC11810778 full-text XML | catalogued No individual claim review recorded Source has a recorded evidence concern. Consult its source page before using the claim. Audit detailsField: Source artifact SHA-256: Hash scope: Hash scope not separately documented; inspect source record |
| Publication status peer_reviewed Source metadata | Nucleotide Transformer: building and evaluating robust foundation models for human genomics (Nature Methods) No field-specific location recorded Version: Nature Methods 22, 287-297 (published online 2024-11-28); PMC11810778 full-text XML | catalogued No individual claim review recorded Source has a recorded evidence concern. Consult its source page before using the claim. Audit detailsField: Source artifact SHA-256: Hash scope: Hash scope not separately documented; inspect source record |
| Retrieved at 2026-10-14T12:48:26Z Source metadata | Nucleotide Transformer: building and evaluating robust foundation models for human genomics (Nature Methods) No field-specific location recorded Version: Nature Methods 22, 287-297 (published online 2024-11-28); PMC11810778 full-text XML | catalogued No individual claim review recorded Source has a recorded evidence concern. Consult its source page before using the claim. Audit detailsField: Source artifact SHA-256: Hash scope: Hash scope not separately documented; inspect source record |
| Url https://doi.org/10.1038/s41592-024-02523-z Source metadata | Nucleotide Transformer: building and evaluating robust foundation models for human genomics (Nature Methods) No field-specific location recorded Version: Nature Methods 22, 287-297 (published online 2024-11-28); PMC11810778 full-text XML | catalogued No individual claim review recorded Source has a recorded evidence concern. Consult its source page before using the claim. Audit detailsField: Source artifact SHA-256: Hash scope: Hash scope not separately documented; inspect source record |
| Version Nature Methods 22, 287-297 (published online 2024-11-28); PMC11810778 full-text XML Source metadata | Nucleotide Transformer: building and evaluating robust foundation models for human genomics (Nature Methods) No field-specific location recorded Version: Nature Methods 22, 287-297 (published online 2024-11-28); PMC11810778 full-text XML | catalogued No individual claim review recorded Source has a recorded evidence concern. Consult its source page before using the claim. Audit detailsField: Source artifact SHA-256: Hash scope: Hash scope not separately documented; inspect source record |
Sources and history
Release 2026-10-22-1b7cf027e4f6 · Record review: source checked
Primary article (Europe PMC full-text XML) retrieved and hashed for the genomic foundation model guide sources pass (issue #138).
Technical metadata and extraction receipts
Stable ID: gfm-guide-20261014-source-nt-nature-methods
- areas
- dna-genomes
- contexts
- research
- artifact url
- https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11810778/fullTextXML
- retrieved at
- 2026-10-14T12:48:26Z
- artifact sha256
- 7c3b78a4f38ef053a08e466222e5662dd9711d91c535512d0b35d459d1fc7249
- url
- https://doi.org/10.1038/s41592-024-02523-z
- doi
- 10.1038/s41592-024-02523-z
- publication status
- peer_reviewed
- licence
- CC-BY-NC-ND-4.0
- version
- Nature Methods 22, 287-297 (published online 2024-11-28); PMC11810778 full-text XML
- media type
- application/xml
- evidence concerns
- source id: gfm-guide-20261014-source-nt-nature-methods; message: Fine-tuning time conflicts within the article. Methods 'Hardware' say all fine-tuning runs used one node of eight A100 GPUs and averaged 20 min (500M-parameter models) and 50 min (2.5B-parameter models); the Discussion says the IA3 technique allows even the largest models to be fine-tuned in under 15 min on a single GPU. Both are recorded; neither is preferred.; source locator: Methods, 'Data preparation > Hardware', paragraph 1; Discussion, paragraph 3; artifact sha256: 7c3b78a4f38ef053a08e466222e5662dd9711d91c535512d0b35d459d1fc7249; reviewed at: 2026-10-14T12:58:25Z; review method: automated-source-review