rewirebio.iobenchmarks
Pipeline

DNABERT-2 117M · frozen pair embeddings

This MFASS pipeline uses a frozen DNABERT-2 encoder with an assay-specific logistic-regression head. It is distinct from the base encoder and from end-to-end fine-tuning.

SourcesMFASS v2 run_dnabert2.py · run_dnabert2.py module documentation and recorded runner configuration

1 evaluation · 3 results

How it worksEvaluated pipeline
Evaluated pipeline1. Assay-oriented sequence pair. Then: 2. Frozen DNABERT-2. Then: 3. Masked mean embeddings. Then: 4. Reference plus difference. Then: 5. Trained logistic headEvaluated pipeline1. Assay-oriented sequence pair. Then: 2. Frozen DNABERT-2. Then: 3. Masked mean embeddings. Then: 4. Reference plus difference. Then: 5. Trained logistic headEvaluated pipeline1. Assay-oriented sequence pair. Then: 2. Frozen DNABERT-2. Then: 3. Masked mean embeddings. Then: 4. Reference plus difference. Then: 5. Trained logistic head

Schematic of the pinned MFASS configuration; this does not generalise to every member of the model family.

SourcesMFASS v2 pinned rewire artifacts · benchmarks/mfass/README.md, DNABERT-2 protocol; results/dnabert2-117m-frozen-pair-logreg.json, config at bee9133b83f3aedaf2bbb9013f1875515845607e

Overview

Model type

Frozen DNA encoder plus fitted logistic-regression pipeline

Biological inputs

Validated 170-bp assay-oriented reference/mutant pairs

Outputs

MFASS splice-disruption probabilities from the fitted logistic head

Access

Pinned public runner; model and tokenizer identities are validated against recorded artifact hashes.

SourcesMFASS v2 run_dnabert2.py · run_dnabert2.py; module documentation, main() configuration and scoring procedure

Source reviewed · Automated source review, 2026-09-16. All specifications and missing details

Model Capabilities & Hardware Requirements

Deployment facts recorded for DNABERT-2 117M · frozen pair embeddings. Fields the record does not carry are marked not recorded.

Parameters117M frozen DNABERT-2 encoder; the logistic head is a…
Minimum VRAMNot recorded
Inference LatencyNot recorded
Context Window170 bases per sequence; tokenizer truncation disabled
AccessPinned public runner; model and tokenizer identities are…
LicenseThe official DNABERT-2-117M checkpoint repository carries…

Evaluated Checkpoint: bee9133b83f3aedaf2bbb9013f1875515845607e

Evaluations and results

1 evaluation · 3 results. Different protocols are not a single leaderboard.

Protocol: MFASS v2 · Dataset: MFASS v2 eligible assay cohort

Sorted by AUROC (higher is better). The best value in each column is highlighted. Decimals are rounded for display; each value links to the printed value and its source.

Tested configurationAUROCAverage precisionPrecision at 100
DNABERT-2 117M · frozen pair embeddingsEvaluation0.550.04510.03
All 3 result rows with coverage, uncertainty and sources
Filter evaluations

Applied filters: All linked evaluations

Exact evaluated configurations and original reported results
Tested configurationProtocol and datasetFindingEvidence and details
Pipeline: DNABERT-2 117M · frozen pair embeddingsProtocol: MFASS v2
Dataset: MFASS v2 eligible assay cohort
0.55 AUROC
unitless · higher

Uncertainty: Not reported by the source: Point estimate; see paired-comparison artifacts.

Coverage: 8324/8324 scored

Rewire evaluation · Rerun recorded
Methods, coverage and source

DNABERT-2 117M · frozen pair embeddings on MFASS v2

Masked mean of frozen last hidden states for 170 bp reference and mutant sequences; concatenate reference and mutant-minus-reference embeddings; fixed balanced L2 logistic head trained only on the MFASS training split.

Aggregation: Not reported

MFASS v2 pinned rewire artifacts · benchmarks/mfass/results/dnabert2-117m-frozen-pair-logreg.json :: auroc
Pipeline: DNABERT-2 117M · frozen pair embeddingsProtocol: MFASS v2
Dataset: MFASS v2 eligible assay cohort
0.0451 Average precision
fraction · higher

Uncertainty: Not reported by the source: Point estimate; see paired-comparison artifacts.

Coverage: 8324/8324 scored

Rewire evaluation · Rerun recorded
Methods, coverage and source

DNABERT-2 117M · frozen pair embeddings on MFASS v2

Masked mean of frozen last hidden states for 170 bp reference and mutant sequences; concatenate reference and mutant-minus-reference embeddings; fixed balanced L2 logistic head trained only on the MFASS training split.

Aggregation: Not reported

MFASS v2 pinned rewire artifacts · benchmarks/mfass/results/dnabert2-117m-frozen-pair-logreg.json :: average_precision
Pipeline: DNABERT-2 117M · frozen pair embeddingsProtocol: MFASS v2
Dataset: MFASS v2 eligible assay cohort
0.03 Precision at 100
fraction · higher

Uncertainty: Not reported by the source: Point estimate; see paired-comparison artifacts.

Coverage: 8324/8324 scored

Rewire evaluation · Rerun recorded
Methods, coverage and source

DNABERT-2 117M · frozen pair embeddings on MFASS v2

Masked mean of frozen last hidden states for 170 bp reference and mutant sequences; concatenate reference and mutant-minus-reference embeddings; fixed balanced L2 logistic head trained only on the MFASS training split.

Aggregation: Not reported

MFASS v2 pinned rewire artifacts · benchmarks/mfass/results/dnabert2-117m-frozen-pair-logreg.json :: precision_at_100

Source checking is not independent reproduction. Release 2026-11-21-adc15377fab2.

Use this pipeline

How it works, versions and access

Underlying model: DNABERT-2. Results on this page belong to this pipeline and its evaluated settings.

How it works

Evaluated procedure

Validated 170-base reference and mutant sequences are embedded separately. Attention-mask means of final hidden states are combined as reference plus mutant-minus-reference. Standardisation and balanced L2 logistic regression are fitted only on the training arm.

SourcesMFASS v2 pinned rewire artifacts · benchmarks/mfass/README.md, DNABERT-2 protocol; results/dnabert2-117m-frozen-pair-logreg.json, config at bee9133b83f3aedaf2bbb9013f1875515845607e
Strengths, limitations and unresolved questions

Strengths and limitations

Strengths supported by sources

  • Separates the utility of a frozen representation from encoder fine-tuning and pins the checkpoint and head settings.
    SourcesMFASS v2 pinned rewire artifacts · benchmarks/mfass/README.md, DNABERT-2 protocol; results/dnabert2-117m-frozen-pair-logreg.json, config at bee9133b83f3aedaf2bbb9013f1875515845607e

Limitations and conditions

  • The result tests this short-context supervised pipeline. It does not establish how all DNABERT-2 adaptations or all foundation models perform.
    SourcesMFASS v2 pinned rewire artifacts · benchmarks/mfass/README.md, DNABERT-2 protocol; results/dnabert2-117m-frozen-pair-logreg.json, config at bee9133b83f3aedaf2bbb9013f1875515845607e
Profile review details

Pinned MFASS runner source rechecked at bee9133b83f3aedaf2bbb9013f1875515845607e. Existing numerical records and historical artifacts remain unchanged; no new run was performed.

Stable record: rewire-model-dnabert2-117m-frozen-pair-logreg

Specifications

Inputs, training, access and other details

Explanatory profile: source reviewed · Automated source review, 2026-09-16. Review applies to the cited claims; unresolved fields are listed below. Numerical results retain their own review status.

Inputs, outputs and configuration
PropertyDescription and evidence
Model typeFrozen DNA encoder plus fitted logistic-regression pipeline
SourcesMFASS v2 run_dnabert2.py · run_dnabert2.py; module documentation, main() configuration and scoring procedure
Architecture / procedureAttention-mask mean pooling of last hidden states for reference and mutant sequences; concatenate reference embedding with mutant-minus-reference, scale features and fit balanced L2 logistic regression with C=0.1 on train rows only.
SourcesMFASS v2 run_dnabert2.py · run_dnabert2.py; module documentation, main() configuration and scoring procedure
Biological inputsValidated 170-bp assay-oriented reference/mutant pairs
SourcesMFASS v2 run_dnabert2.py · run_dnabert2.py; module documentation, main() configuration and scoring procedure
OutputsMFASS splice-disruption probabilities from the fitted logistic head
SourcesMFASS v2 run_dnabert2.py · run_dnabert2.py; module documentation, main() configuration and scoring procedure
Parameters117M frozen DNABERT-2 encoder; the logistic head is a separate fitted component.
SourcesMFASS v2 run_dnabert2.py · run_dnabert2.py; module documentation, main() configuration and scoring procedure
Training data / fittingThe encoder remains frozen; StandardScaler and logistic regression fit split-v2 training variants without test tuning.
SourcesMFASS v2 run_dnabert2.py · run_dnabert2.py; module documentation, main() configuration and scoring procedure
Context limits170 bases per sequence; tokenizer truncation disabled.
SourcesMFASS v2 run_dnabert2.py · run_dnabert2.py; module documentation, main() configuration and scoring procedure
Known versions / configurationzhihan1996/DNABERT-2-117M; runner revision bee9133b83f3aedaf2bbb9013f1875515845607e
SourcesMFASS v2 run_dnabert2.py · run_dnabert2.py; module documentation, main() configuration and scoring procedure
AccessPinned public runner; model and tokenizer identities are validated against recorded artifact hashes.
SourcesMFASS v2 run_dnabert2.py · run_dnabert2.py; module documentation, main() configuration and scoring procedure
Code licenceMIT for the Rewire runner at the reviewed revision. Upstream DNABERT-2: Apache 2.0. These are separate component licences.
Sources (2)Rewire benchmark runner licence; MAGICS-LAB/DNABERT_2 LICENSE · LICENSE; complete MIT grant; upstream LICENSE
Weights licenceThe official DNABERT-2-117M checkpoint repository carries Apache 2.0. The separately fitted logistic-regression artifact is not assigned an additional inferred licence.
Sourceshuggingface.co/zhihan1996/DNABERT-2-117M LICENSE · LICENSE in official checkpoint repository

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.

21 evidence rows matching the loaded filters

Claims, original sources and review scope · Release 2026-11-21-adc15377fab2
Property and statementOriginal source and locationReview and provenance
Diagram caption
Schematic of the pinned MFASS configuration; this does not generalise to every member of the model family.
Individual claims
MFASS v2 pinned rewire artifacts

Original source ↗

benchmarks/mfass/README.md, DNABERT-2 protocol; results/dnabert2-117m-frozen-pair-logreg.json, config at bee9133b83f3aedaf2bbb9013f1875515845607e

Version: bee9133b83f3aedaf2bbb9013f1875515845607e
Retrieved: 2026-09-16T10:50:02Z

source checked

automated source review · 2026-09-16

Audit details

Pinned MFASS runner source rechecked at bee9133b83f3aedaf2bbb9013f1875515845607e. Existing numerical records and historical artifacts remain unchanged; no new run was performed.

Field: attributes.profile.diagram.caption

Source artifact SHA-256: a3af693afc070b39b334e359beed7f9affd76f7456a4f9264e2d1e9dab26111d

Hash scope: Local imported data/benchmark-runs/mfass-v2.json; not the upstream repository or result-file hash

Diagram steps
  • Assay-oriented sequence pair
  • Frozen DNABERT-2
  • Masked mean embeddings
  • Reference plus difference
  • Trained logistic head
Individual claims
MFASS v2 pinned rewire artifacts

Original source ↗

benchmarks/mfass/README.md, DNABERT-2 protocol; results/dnabert2-117m-frozen-pair-logreg.json, config at bee9133b83f3aedaf2bbb9013f1875515845607e

Version: bee9133b83f3aedaf2bbb9013f1875515845607e
Retrieved: 2026-09-16T10:50:02Z

source checked

automated source review · 2026-09-16

Audit details

Pinned MFASS runner source rechecked at bee9133b83f3aedaf2bbb9013f1875515845607e. Existing numerical records and historical artifacts remain unchanged; no new run was performed.

Field: attributes.profile.diagram.steps

Source artifact SHA-256: a3af693afc070b39b334e359beed7f9affd76f7456a4f9264e2d1e9dab26111d

Hash scope: Local imported data/benchmark-runs/mfass-v2.json; not the upstream repository or result-file hash

Diagram title
Evaluated pipeline
Individual claims
MFASS v2 pinned rewire artifacts

Original source ↗

benchmarks/mfass/README.md, DNABERT-2 protocol; results/dnabert2-117m-frozen-pair-logreg.json, config at bee9133b83f3aedaf2bbb9013f1875515845607e

Version: bee9133b83f3aedaf2bbb9013f1875515845607e
Retrieved: 2026-09-16T10:50:02Z

source checked

automated source review · 2026-09-16

Audit details

Pinned MFASS runner source rechecked at bee9133b83f3aedaf2bbb9013f1875515845607e. Existing numerical records and historical artifacts remain unchanged; no new run was performed.

Field: attributes.profile.diagram.title

Source artifact SHA-256: a3af693afc070b39b334e359beed7f9affd76f7456a4f9264e2d1e9dab26111d

Hash scope: Local imported data/benchmark-runs/mfass-v2.json; not the upstream repository or result-file hash

Model type
Frozen DNA encoder plus fitted logistic-regression pipeline
Individual claims
MFASS v2 run_dnabert2.py

Original source ↗

run_dnabert2.py; module documentation, main() configuration and scoring procedure

Version: bee9133b83f3aedaf2bbb9013f1875515845607e
Retrieved: 2026-09-16T20:47:15.959417+00:00

source checked

automated source review · 2026-09-16

Audit details

Pinned MFASS runner source rechecked at bee9133b83f3aedaf2bbb9013f1875515845607e. Existing numerical records and historical artifacts remain unchanged; no new run was performed.

Field: attributes.profile.facts.0.value

Source artifact SHA-256: b7ebb4324e22cb421482181c52722c64cac5fec7c43161c082fc34e9920c4a85

Hash scope: Hash scope not separately documented; inspect source record

Inspected artifact

Architecture / procedure
Attention-mask mean pooling of last hidden states for reference and mutant sequences; concatenate reference embedding with mutant-minus-reference, scale features and fit balanced L2 logistic regression with C=0.1 on train rows only.
Individual claims
MFASS v2 run_dnabert2.py

Original source ↗

run_dnabert2.py; module documentation, main() configuration and scoring procedure

Version: bee9133b83f3aedaf2bbb9013f1875515845607e
Retrieved: 2026-09-16T20:47:15.959417+00:00

source checked

automated source review · 2026-09-16

Audit details

Pinned MFASS runner source rechecked at bee9133b83f3aedaf2bbb9013f1875515845607e. Existing numerical records and historical artifacts remain unchanged; no new run was performed.

Field: attributes.profile.facts.1.value

Source artifact SHA-256: b7ebb4324e22cb421482181c52722c64cac5fec7c43161c082fc34e9920c4a85

Hash scope: Hash scope not separately documented; inspect source record

Inspected artifact

Weights licence
The official DNABERT-2-117M checkpoint repository carries Apache 2.0. The separately fitted logistic-regression artifact is not assigned an additional inferred licence.
Individual claims
huggingface.co/zhihan1996/DNABERT-2-117M LICENSE

Original source ↗

LICENSE in official checkpoint repository

Version: 7bce263b15377fc15361f52cfab88f8b586abda0
Retrieved: 2026-09-16T19:46:20.874956+00:00

source checked

automated source review · 2026-09-16

Audit details

Pinned MFASS runner source rechecked at bee9133b83f3aedaf2bbb9013f1875515845607e. Existing numerical records and historical artifacts remain unchanged; no new run was performed.

Field: attributes.profile.facts.10.value

Source artifact SHA-256: c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4

Hash scope: Hash scope not separately documented; inspect source record

Inspected artifact

Biological inputs
Validated 170-bp assay-oriented reference/mutant pairs
Individual claims
MFASS v2 run_dnabert2.py

Original source ↗

run_dnabert2.py; module documentation, main() configuration and scoring procedure

Version: bee9133b83f3aedaf2bbb9013f1875515845607e
Retrieved: 2026-09-16T20:47:15.959417+00:00

source checked

automated source review · 2026-09-16

Audit details

Pinned MFASS runner source rechecked at bee9133b83f3aedaf2bbb9013f1875515845607e. Existing numerical records and historical artifacts remain unchanged; no new run was performed.

Field: attributes.profile.facts.2.value

Source artifact SHA-256: b7ebb4324e22cb421482181c52722c64cac5fec7c43161c082fc34e9920c4a85

Hash scope: Hash scope not separately documented; inspect source record

Inspected artifact

Outputs
MFASS splice-disruption probabilities from the fitted logistic head
Individual claims
MFASS v2 run_dnabert2.py

Original source ↗

run_dnabert2.py; module documentation, main() configuration and scoring procedure

Version: bee9133b83f3aedaf2bbb9013f1875515845607e
Retrieved: 2026-09-16T20:47:15.959417+00:00

source checked

automated source review · 2026-09-16

Audit details

Pinned MFASS runner source rechecked at bee9133b83f3aedaf2bbb9013f1875515845607e. Existing numerical records and historical artifacts remain unchanged; no new run was performed.

Field: attributes.profile.facts.3.value

Source artifact SHA-256: b7ebb4324e22cb421482181c52722c64cac5fec7c43161c082fc34e9920c4a85

Hash scope: Hash scope not separately documented; inspect source record

Inspected artifact

Parameters
117M frozen DNABERT-2 encoder; the logistic head is a separate fitted component.
Individual claims
MFASS v2 run_dnabert2.py

Original source ↗

run_dnabert2.py; module documentation, main() configuration and scoring procedure

Version: bee9133b83f3aedaf2bbb9013f1875515845607e
Retrieved: 2026-09-16T20:47:15.959417+00:00

source checked

automated source review · 2026-09-16

Audit details

Pinned MFASS runner source rechecked at bee9133b83f3aedaf2bbb9013f1875515845607e. Existing numerical records and historical artifacts remain unchanged; no new run was performed.

Field: attributes.profile.facts.4.value

Source artifact SHA-256: b7ebb4324e22cb421482181c52722c64cac5fec7c43161c082fc34e9920c4a85

Hash scope: Hash scope not separately documented; inspect source record

Inspected artifact

Training data / fitting
The encoder remains frozen; StandardScaler and logistic regression fit split-v2 training variants without test tuning.
Individual claims
MFASS v2 run_dnabert2.py

Original source ↗

run_dnabert2.py; module documentation, main() configuration and scoring procedure

Version: bee9133b83f3aedaf2bbb9013f1875515845607e
Retrieved: 2026-09-16T20:47:15.959417+00:00

source checked

automated source review · 2026-09-16

Audit details

Pinned MFASS runner source rechecked at bee9133b83f3aedaf2bbb9013f1875515845607e. Existing numerical records and historical artifacts remain unchanged; no new run was performed.

Field: attributes.profile.facts.5.value

Source artifact SHA-256: b7ebb4324e22cb421482181c52722c64cac5fec7c43161c082fc34e9920c4a85

Hash scope: Hash scope not separately documented; inspect source record

Inspected artifact

How to read the evidence

Sources and history

Release 2026-11-21-adc15377fab2 · Record review: source checked

5 source records and release historyDownload this release (gzip)
Technical metadata and record identifiers

Technical metadata for this record is tracked under stable ID rewire-model-dnabert2-117m-frozen-pair-logreg.

Related records

How to cite

DNABERT-2 117M · frozen pair embeddings. Rewire biological model benchmark database, release 2026-11-21-adc15377fab2. https://benchmarks.rewirebio.io/database/pipeline/rewire-model-dnabert2-117m-frozen-pair-logreg/

Release 2026-11-21-adc15377fab2. Rankings and values change between releases, so keep the release id in the citation. Cite the original publication as well for any value you reuse: each result links its source.

The compilation and annotations made by Rewire in this release are offered under CC BY 4.0; values and text taken from cited sources stay under the terms of those sources and are not covered by this licence. Data licence

Suggest a correction