Relevance of Fine-tuning and evaluation wall time, 5 epochs on the Enformer eQTL dataset (Hegde et al. 2025 Table A1) to "Select an execution workflow for large-scale diagnostic genomics"
- Use case
- Select an execution workflow for large-scale diagnostic genomics
- Assessed by
- Fine-tuning and evaluation wall time, 5 epochs on the Enformer eQTL dataset (Hegde et al. 2025 Table A1)
- Relevance
- Proxy evidence
- Endpoint
- Fine-tuning wall time (hours) for 5 epochs with LoRA (r=1) on the eQTL training split (89,060 variants in Table 1; the Methods text says 88,717) and evaluation wall time (minutes) on the 8,862-variant test split, Nucleotide Transformer 500M human-ref, one row per seed (three seeds)
Proxy. Wall time to adapt a genomic foundation model with LoRA to a human non-coding variant classification task and to score a held-out set of 8,862 human variants, measured by a group that did not develop the model, with three seeds. It informs the compute a team needs to fine-tune and run such a model on its own hardware, but it is a research benchmark rather than a diagnostic workflow, and the table does not name its hardware.
1 reviewed evaluation
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 |
|---|---|---|
Citation locations1 values
| Reaping the Fruits of LLM Pruning: Towards Small Language Models for Efficient Non-Coding Variant Effect Prediction gfm-guide-20261014-source-hegde2025: Table A1; Methods paragraph 3; Table 1 (train and test split sizes); Appendix A.3 Version: Genes 16(11):1358, 2025; PMC12652071 full-text XML | not individually reviewed No individual claim review recorded Audit detailsField: Source artifact SHA-256: Hash scope: Hash scope not separately documented; inspect source record |
| Comparison group hegde2025-eqtl-finetune Context-only references | Reaping the Fruits of LLM Pruning: Towards Small Language Models for Efficient Non-Coding Variant Effect Prediction gfm-guide-20261014-source-hegde2025: Table A1; Methods paragraph 3; Table 1 (train and test split sizes); Appendix A.3 Version: Genes 16(11):1358, 2025; PMC12652071 full-text XML | not individually reviewed No individual claim review recorded Audit detailsField: Source artifact SHA-256: Hash scope: Hash scope not separately documented; inspect source record |
| Comparison title Nucleotide Transformer 500M fine-tuning and evaluation time on the Enformer eQTL task (Hegde et al. 2025) Context-only references | Reaping the Fruits of LLM Pruning: Towards Small Language Models for Efficient Non-Coding Variant Effect Prediction gfm-guide-20261014-source-hegde2025: Table A1; Methods paragraph 3; Table 1 (train and test split sizes); Appendix A.3 Version: Genes 16(11):1358, 2025; PMC12652071 full-text XML | not individually reviewed No individual claim review recorded Audit detailsField: Source artifact SHA-256: Hash scope: Hash scope not separately documented; inspect source record |
Constraints4 values
| Reaping the Fruits of LLM Pruning: Towards Small Language Models for Efficient Non-Coding Variant Effect Prediction gfm-guide-20261014-source-hegde2025: Table A1; Methods paragraph 3; Table 1 (train and test split sizes); Appendix A.3 Version: Genes 16(11):1358, 2025; PMC12652071 full-text XML | not individually reviewed No individual claim review recorded Audit detailsField: Source artifact SHA-256: Hash scope: Hash scope not separately documented; inspect source record |
| Endpoint Fine-tuning wall time (hours) for 5 epochs with LoRA (r=1) on the eQTL training split (89,060 variants in Table 1; the Methods text says 88,717) and evaluation wall time (minutes) on the 8,862-variant test split, Nucleotide Transformer 500M human-ref, one row per seed (three seeds) Context-only references | Reaping the Fruits of LLM Pruning: Towards Small Language Models for Efficient Non-Coding Variant Effect Prediction gfm-guide-20261014-source-hegde2025: Table A1; Methods paragraph 3; Table 1 (train and test split sizes); Appendix A.3 Version: Genes 16(11):1358, 2025; PMC12652071 full-text XML | not individually reviewed No individual claim review recorded Audit detailsField: Source artifact SHA-256: Hash scope: Hash scope not separately documented; inspect source record |
| Claimed field links:assessed_by:gfm-guide-20261014-protocol-hegde-eqtl-finetune-5-epochs-time Context-only references | Reaping the Fruits of LLM Pruning: Towards Small Language Models for Efficient Non-Coding Variant Effect Prediction gfm-guide-20261014-source-hegde2025: Table A1; Methods paragraph 3; Table 1 (train and test split sizes); Appendix A.3 Version: Genes 16(11):1358, 2025; PMC12652071 full-text XML | not individually reviewed No individual claim review recorded Audit detailsField: Source artifact SHA-256: Hash scope: Hash scope not separately documented; inspect source record |
| Headline metric runtime Context-only references | Reaping the Fruits of LLM Pruning: Towards Small Language Models for Efficient Non-Coding Variant Effect Prediction gfm-guide-20261014-source-hegde2025: Table A1; Methods paragraph 3; Table 1 (train and test split sizes); Appendix A.3 Version: Genes 16(11):1358, 2025; PMC12652071 full-text XML | not individually reviewed No individual claim review recorded Audit detailsField: Source artifact SHA-256: Hash scope: Hash scope not separately documented; inspect source record |
Limitations6 values
| Reaping the Fruits of LLM Pruning: Towards Small Language Models for Efficient Non-Coding Variant Effect Prediction gfm-guide-20261014-source-hegde2025: Table A1; Methods paragraph 3; Table 1 (train and test split sizes); Appendix A.3 Version: Genes 16(11):1358, 2025; PMC12652071 full-text XML | not individually reviewed No individual claim review recorded Audit detailsField: Source artifact SHA-256: Hash scope: Hash scope not separately documented; inspect source record |
Pinned records10 values
| Reaping the Fruits of LLM Pruning: Towards Small Language Models for Efficient Non-Coding Variant Effect Prediction gfm-guide-20261014-source-hegde2025: Table A1; Methods paragraph 3; Table 1 (train and test split sizes); Appendix A.3 Version: Genes 16(11):1358, 2025; PMC12652071 full-text XML | not individually reviewed No individual claim review recorded Audit detailsField: Source artifact SHA-256: Hash scope: Hash scope not separately documented; inspect source record |
| Rationale Proxy. Wall time to adapt a genomic foundation model with LoRA to a human non-coding variant classification task and to score a held-out set of 8,862 human variants, measured by a group that did not develop the model, with three seeds. It informs the compute a team needs to fine-tune and run such a model on its own hardware, but it is a research benchmark rather than a diagnostic workflow, and the table does not name its hardware. Context-only references | Reaping the Fruits of LLM Pruning: Towards Small Language Models for Efficient Non-Coding Variant Effect Prediction gfm-guide-20261014-source-hegde2025: Table A1; Methods paragraph 3; Table 1 (train and test split sizes); Appendix A.3 Version: Genes 16(11):1358, 2025; PMC12652071 full-text XML | not individually reviewed No individual claim review recorded 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
1 source record and release history
- Reaping the Fruits of LLM Pruning: Towards Small Language Models for Efficient Non-Coding Variant Effect Prediction · Original source · Genes 16(11):1358, 2025; PMC12652071 full-text XML
Technical metadata and extraction receipts
Stable ID: use-case-mapping-gfm-guide-20261014-hegde2025-eqtl-5-epochs
- field
- links:assessed_by:gfm-guide-20261014-protocol-hegde-eqtl-finetune-5-epochs-time
- value
- gfm-guide-20261014-protocol-hegde-eqtl-finetune-5-epochs-time
- relevance
- proxy
- endpoint
- Fine-tuning wall time (hours) for 5 epochs with LoRA (r=1) on the eQTL training split (89,060 variants in Table 1; the Methods text says 88,717) and evaluation wall time (minutes) on the 8,862-variant test split, Nucleotide Transformer 500M human-ref, one row per seed (three seeds)
- rationale
- Proxy. Wall time to adapt a genomic foundation model with LoRA to a human non-coding variant classification task and to score a held-out set of 8,862 human variants, measured by a group that did not develop the model, with three seeds. It informs the compute a team needs to fine-tune and run such a model on its own hardware, but it is a research benchmark rather than a diagnostic workflow, and the table does not name its hardware.
- constraints
- Inspect every linked evaluation's source locator before citing a result.; Do not combine this mapping's evaluations with any other protocol's results; hardware, inputs, software versions and timing boundaries differ between sources.; Give the hardware as the Methods state it for the Nucleotide Transformer (2 × NVIDIA TITAN Xp), and say that Table A1 names none.; Do not read the three seeds as run-to-run concordance of variant calls; they repeat the whole fine-tuning and evaluation, and Table A1 prints identical aggregate accuracy, AUROC and F1, not per-variant agreement.
- limitations
- Table A1 names no hardware. The Methods give 2 × NVIDIA TITAN Xp for training the Nucleotide Transformer to convergence and say the two models were trained on different machines, so these runs are read as on the same GPUs; the TITAN Xp is older than the GPUs in the other linked studies.; Three seeds; values are printed per seed and no spread is printed.; Fine-tuning is LoRA through peft with r=1 (Appendix A.3), not full fine-tuning.; Fixed 5-epoch budget, not training to convergence (the convergence time is a separate protocol).; The training split is 89,060 variants in Table 1 but 88,717 in the Methods text.; Evaluation time covers the whole test split; per-variant throughput and peak memory are not printed.
- revision
- 1
- reason
- Add relevance judgements for the execution and compute protocols of the genomic foundation model guide sources batch (data/omics/gfm-guide-sources-20261014, issue #162), independently reviewed in PR #169.
- citation locators
- source id: gfm-guide-20261014-source-hegde2025; locator: Table A1; Methods paragraph 3; Table 1 (train and test split sizes); Appendix A.3
- source locator
- gfm-guide-20261014-source-hegde2025: Table A1; Methods paragraph 3; Table 1 (train and test split sizes); Appendix A.3
- comparison group
- hegde2025-eqtl-finetune
- comparison title
- Nucleotide Transformer 500M fine-tuning and evaluation time on the Enformer eQTL task (Hegde et al. 2025)
- headline metric
- runtime
- reviewed evaluations
- gfm-guide-20261014-eval-hegde-nt-500m-human-ref-5-epochs
- review
- method: source-hash-verification; independent-cell-check; scope-review; reviewer: claude; reviewer note: Claude (Opus 5.5) review agent review-169, independent of the agent that drafted these judgements for issue #162; no human review claimed; reviewed at: 2026-10-14T19:00:00Z; note: Independent relevance review: grade, endpoint, rationale, constraints, limitations and locators checked against the use case and the source text; every value of the listed evaluations checked against the source. No execution, reproduction, qualified human scientific review or clinical validation. See docs/reviews/use-cases/diagnostic-genomics-model-execution-judgements-2026-10-14.md. Reviewer corrected the endpoint, constraints and limitations (LoRA, split-size conflict, hardware attribution, meaning of the seeds).
- pins
- record id: gfm-guide-20261014-eval-hegde-nt-500m-human-ref-5-epochs; fields: status; links; attributes.origin; attributes.comparison; sha256: 48fcaf7dadc2352f4f8686550e972c13075812ce01feb90d506f7b6893a4553f; record id: gfm-guide-20261014-protocol-hegde-eqtl-finetune-5-epochs-time; fields: status; name; description; facets; source_ids; links; attributes; sha256: 90ce4997552ed8215045a4388cf5302aff15de103f0986354403ed505b98dcf0; record id: gfm-guide-20261014-result-hegde-nt-5-epochs-evaluation-time-seed-41; fields: status; attributes.metric; attributes.metric_qualifier; attributes.printed_value; attributes.numeric_value; sha256: 46d13ab60cdbed1905520751bfcf426518be59e0f4a552a015e51632e7de6dd7; record id: gfm-guide-20261014-result-hegde-nt-5-epochs-evaluation-time-seed-42; fields: status; attributes.metric; attributes.metric_qualifier; attributes.printed_value; attributes.numeric_value; sha256: 848d6d71ff7fbbe20a7fa995da5a47cd80d0655fd46aa93210f3a4d8ec8a377b; record id: gfm-guide-20261014-result-hegde-nt-5-epochs-evaluation-time-seed-43; fields: status; attributes.metric; attributes.metric_qualifier; attributes.printed_value; attributes.numeric_value; sha256: 46d13ab60cdbed1905520751bfcf426518be59e0f4a552a015e51632e7de6dd7; record id: gfm-guide-20261014-result-hegde-nt-5-epochs-training-time-seed-41; fields: status; attributes.metric; attributes.metric_qualifier; attributes.printed_value; attributes.numeric_value; sha256: a947621929c9574f651d200224e50474470c0ce4c5b30b234cfd089153bf9292; record id: gfm-guide-20261014-result-hegde-nt-5-epochs-training-time-seed-42; fields: status; attributes.metric; attributes.metric_qualifier; attributes.printed_value; attributes.numeric_value; sha256: 2e59a220b0aa7f073d6b22db1b696ef10f4d8da608b3743729caab508faf4090; record id: gfm-guide-20261014-result-hegde-nt-5-epochs-training-time-seed-43; fields: status; attributes.metric; attributes.metric_qualifier; attributes.printed_value; attributes.numeric_value; sha256: a3758879127318c019da9c2d1107d738f2577833b994386acb1597a2b45d7a6b; record id: gfm-guide-20261014-source-hegde2025; fields: status; attributes.artifact_sha256; sha256: 4ec6ee59ca3c282cc9f37731f592837db20813b4ad179f1960acfe6782221a4e; record id: use-case-diagnostic-genomics-model-execution; fields: status; name; description; facets; attributes.question; attributes.decision; attributes.inputs; attributes.output; attributes.exclusions; attributes.setting; attributes.clinical_scope; attributes.intended_users; sha256: 1c2a12431556cc6581f5aab019cf91576d2393387b88a8a2bea35ea2a3ee4a70