Task

protein localization classification

Protein-localization classification evaluates frozen microscopy-image representations using Human Protein Atlas tasks.

Sources

Cell-DINO: Self-supervised image-based embeddings for cell fluorescent microscopy · Methods: Training and evaluation protocol on HPA datasets; Kaggle evaluation protocol; cached text lines 46, 86–94; task metric definitions and corresponding results table

1 evaluation · 1 metric row

At a glance

Inputs, training, access and other details

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

Data, procedure and scoring
Property	Description and evidence
Datasets	HPA field-of-view and single-cell image tasks. Sources Cell-DINO: Self-supervised image-based embeddings for cell fluorescent microscopy · Methods: Training and evaluation protocol on HPA datasets; Kaggle evaluation protocol; cached text lines 46, 86–94; task metric definitions and corresponding results table
Splits	The Kaggle field-of-view evaluation uses five folds to choose and average class thresholds. Sources Cell-DINO: Self-supervised image-based embeddings for cell fluorescent microscopy · Methods: Training and evaluation protocol on HPA datasets; Kaggle evaluation protocol; cached text lines 46, 86–94; task metric definitions and corresponding results table
Metrics	F1 for protein-localization classification; cell-line accuracy and Cell Painting metrics are separate tasks. Sources Cell-DINO: Self-supervised image-based embeddings for cell fluorescent microscopy · Methods: Training and evaluation protocol on HPA datasets; Kaggle evaluation protocol; cached text lines 46, 86–94; task metric definitions and corresponding results table
Baselines	The paper includes nearest-neighbour, linear and MLP-based evaluations; these are distinct classifier settings. Sources Cell-DINO: Self-supervised image-based embeddings for cell fluorescent microscopy · Methods: Training and evaluation protocol on HPA datasets; Kaggle evaluation protocol; cached text lines 46, 86–94; task metric definitions and corresponding results table
Leakage controls	Routine HPA train/validation analysis and Kaggle public-test generalization are different settings; per-class thresholds are selected on validation folds rather than inferred from the public score. Sources Cell-DINO: Self-supervised image-based embeddings for cell fluorescent microscopy · Methods: Training and evaluation protocol on HPA datasets; Kaggle evaluation protocol; cached text lines 46, 86–94; task metric definitions and corresponding results table
Uncertainty	The protein-localization tables report point F1 or challenge scores. The HPA evaluation methods and S1 Text do not define repeated-seed confidence intervals or an uncertainty estimator for these values; comparisons between protocols retain their distinct classifiers and label access. · Not reported in inspected sources Sources (2) Cell-DINO: Self-supervised image-based embeddings for cell fluorescent microscopy; cell-dino-2025__pcbi.1013828.s001.pdf · Methods: Training and evaluation protocol on HPA datasets; Tables 1–3 and 6; S1 Text
Entity type	Paper-specific computational evaluation protocol. Sources Cell-DINO: Self-supervised image-based embeddings for cell fluorescent microscopy · Methods: Training and evaluation protocol on HPA datasets; Kaggle evaluation protocol; cached text lines 46, 86–94; task metric definitions and corresponding results table
Organisms	Human Protein Atlas cells. Sources Cell-DINO: Self-supervised image-based embeddings for cell fluorescent microscopy · Methods: Training and evaluation protocol on HPA datasets; Kaggle evaluation protocol; cached text lines 46, 86–94; task metric definitions and corresponding results table
Assays	Microscopy images with protein-localization labels. Sources Cell-DINO: Self-supervised image-based embeddings for cell fluorescent microscopy · Methods: Training and evaluation protocol on HPA datasets; Kaggle evaluation protocol; cached text lines 46, 86–94; task metric definitions and corresponding results table
Allowed inputs	Field-of-view or single-cell images, depending on the task setting. Sources Cell-DINO: Self-supervised image-based embeddings for cell fluorescent microscopy · Methods: Training and evaluation protocol on HPA datasets; Kaggle evaluation protocol; cached text lines 46, 86–94; task metric definitions and corresponding results table
Adaptation	Frozen-feature nearest-neighbour/linear/MLP evaluation regimes are distinct; threshold selection uses folds. Sources Cell-DINO: Self-supervised image-based embeddings for cell fluorescent microscopy · Methods: Training and evaluation protocol on HPA datasets; Kaggle evaluation protocol; cached text lines 46, 86–94; task metric definitions and corresponding results table

How it works

How it worksComputational evaluation flow

Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol.

Sources

Evaluation methodology

HPA field-of-view and single-cell image tasks. The Kaggle field-of-view evaluation uses five folds to choose and average class thresholds. F1 for protein-localization classification; cell-line accuracy and Cell Painting metrics are separate tasks. The paper includes nearest-neighbour, linear and MLP-based evaluations; these are distinct classifier settings. Routine HPA train/validation analysis and Kaggle public-test generalization are different settings; per-class thresholds are selected on validation folds rather than inferred from the public score.

Sources

Recorded evaluations

Each evaluation records what was tested and under which conditions.

Cell-DINO ViT-L: protein localization classification

Tested entities and results

Release 2026-09-17-d277315f7d76 · 1 evaluation · 1 metric row. Different protocols are not a single leaderboard.

Evaluation setupMetricEvidence origin

Results grouped by the exact reported evaluation
Metric and finding	Coverage and uncertainty	Evidence
Cell-DINO ViT-L: protein localization classification Configuration: Cell-DINO ViT-LTask: protein localization classificationDataset: HPA-FoV Self-supervised microscopy embedding pre-trained on HPA-FoV; downstream protein-localization classifier. Dataset-specific pretraining; the paper does not claim a general-purpose foundation model that generalizes beyond these benchmarks. Author-reported evaluation · Evaluation metadata: needs review
65.5% F1 Unit: percent · Direction: unknown	Uncertainty: not reported in legacy extract Scored: Not reported · Eligible: Not reported	source checkedCell-DINO: Self-supervised image-based embeddings for cell fluorescent microscopy · Table 2, HPA-FoV section, Cell-DINO row, PL column Source checking is not independent reproduction.

Papers and result coverage

Last literature check: 2026-09-17. Primary-paper discovery and source inspection. Source-checked results are not independently reproduced experiments.

Paper or primary resource	Version	Reference
Cell-DINO: Self-supervised image-based embeddings for cell fluorescent microscopy	version of record	Read source DOI: 10.1371/journal.pcbi.1013828

What is still missing

Table 2 architectures and pretraining datasets differ; HPA-FoV and HPA-SC are separate protocols.
Do not attach Cell Painting MoA AUPRC to protein localization.
Table 6 ensembling is five MLPs with threshold/prediction averaging; not the base frozen model alone.

Search and extraction details

primary comparison table screened

Searches

"PMC12826486"

Evidence locations

Tables 2 and 6
Methods: Kaggle evaluation protocol

Strengths and limitations

Strengths and considerations

No source-reviewed explanatory claims are recorded here yet.

Limitations and conditions

Field-of-view challenge ensembling, single-cell classification and nearest-neighbor evaluation use different classifiers and protocols. Their scores should not be merged into one interchangeable result.
Sources (2)
Cell-DINO: Self-supervised image-based embeddings for cell fluorescent microscopy; cell-dino-2025__pcbi.1013828.s001.pdf · Methods: Training and evaluation protocol on HPA datasets; Tables 1–3 and 6; S1 Text

Profile review details

Targeted full-paper and supplement review of the outstanding task fields, with original dataset metadata checked where accessible. Source-scoped omissions are explicit; no independent benchmark reproduction or numerical-result change.

Stable record: reported-task-7621fa1be55362

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.

19 evidence rows matching the loaded filters

Claims, original sources and review scope · Release 2026-09-17-d277315f7d76
Property and statement	Original source and location	Review and provenance
Diagram caption Conceptual summary of the cited evaluation; exact task configuration and source version remain part of the protocol. Individual claims	Cell-DINO: Self-supervised image-based embeddings for cell fluorescent microscopy Original source ↗ Methods: Training and evaluation protocol on HPA datasets; Kaggle evaluation protocol; cached text lines 46, 86–94; task metric definitions and corresponding results table Version: version of record Retrieved: 2026-09-16T10:41:06Z	source checked automated source review · 2026-09-16 Audit details Targeted full-paper and supplement review of the outstanding task fields, with original dataset metadata checked where accessible. Source-scoped omissions are explicit; no independent benchmark reproduction or numerical-result change. Field: `attributes.profile.diagram.caption` Source artifact SHA-256: `12a53a78c70b3033c3351cf7afd4da42ebc98bb3281308f07e71e5baffc153a0` Hash scope: Hash scope not separately documented; inspect source record Inspected artifact
Diagram steps ["Input: Field-of-view or single-cell images, depending on the task setting.","Evaluation: The Kaggle field-of-view evaluation uses five folds to choose and average class thresholds.","Readout: F1 for protein-localization classification; cell-line accuracy and Cell Painting metrics are separate tasks."] Individual claims	Cell-DINO: Self-supervised image-based embeddings for cell fluorescent microscopy Original source ↗ Methods: Training and evaluation protocol on HPA datasets; Kaggle evaluation protocol; cached text lines 46, 86–94; task metric definitions and corresponding results table Version: version of record Retrieved: 2026-09-16T10:41:06Z	source checked automated source review · 2026-09-16 Audit details Targeted full-paper and supplement review of the outstanding task fields, with original dataset metadata checked where accessible. Source-scoped omissions are explicit; no independent benchmark reproduction or numerical-result change. Field: `attributes.profile.diagram.steps` Source artifact SHA-256: `12a53a78c70b3033c3351cf7afd4da42ebc98bb3281308f07e71e5baffc153a0` Hash scope: Hash scope not separately documented; inspect source record Inspected artifact
Diagram title Computational evaluation flow Individual claims	Cell-DINO: Self-supervised image-based embeddings for cell fluorescent microscopy Original source ↗ Methods: Training and evaluation protocol on HPA datasets; Kaggle evaluation protocol; cached text lines 46, 86–94; task metric definitions and corresponding results table Version: version of record Retrieved: 2026-09-16T10:41:06Z	source checked automated source review · 2026-09-16 Audit details Targeted full-paper and supplement review of the outstanding task fields, with original dataset metadata checked where accessible. Source-scoped omissions are explicit; no independent benchmark reproduction or numerical-result change. Field: `attributes.profile.diagram.title` Source artifact SHA-256: `12a53a78c70b3033c3351cf7afd4da42ebc98bb3281308f07e71e5baffc153a0` Hash scope: Hash scope not separately documented; inspect source record Inspected artifact
Datasets HPA field-of-view and single-cell image tasks. Individual claims	Cell-DINO: Self-supervised image-based embeddings for cell fluorescent microscopy Original source ↗ Methods: Training and evaluation protocol on HPA datasets; Kaggle evaluation protocol; cached text lines 46, 86–94; task metric definitions and corresponding results table Version: version of record Retrieved: 2026-09-16T10:41:06Z	source checked automated source review · 2026-09-16 Audit details Targeted full-paper and supplement review of the outstanding task fields, with original dataset metadata checked where accessible. Source-scoped omissions are explicit; no independent benchmark reproduction or numerical-result change. Field: `attributes.profile.facts.0.value` Source artifact SHA-256: `12a53a78c70b3033c3351cf7afd4da42ebc98bb3281308f07e71e5baffc153a0` Hash scope: Hash scope not separately documented; inspect source record Inspected artifact
Splits The Kaggle field-of-view evaluation uses five folds to choose and average class thresholds. Individual claims	Cell-DINO: Self-supervised image-based embeddings for cell fluorescent microscopy Original source ↗ Methods: Training and evaluation protocol on HPA datasets; Kaggle evaluation protocol; cached text lines 46, 86–94; task metric definitions and corresponding results table Version: version of record Retrieved: 2026-09-16T10:41:06Z	source checked automated source review · 2026-09-16 Audit details Targeted full-paper and supplement review of the outstanding task fields, with original dataset metadata checked where accessible. Source-scoped omissions are explicit; no independent benchmark reproduction or numerical-result change. Field: `attributes.profile.facts.1.value` Source artifact SHA-256: `12a53a78c70b3033c3351cf7afd4da42ebc98bb3281308f07e71e5baffc153a0` Hash scope: Hash scope not separately documented; inspect source record Inspected artifact
Adaptation Frozen-feature nearest-neighbour/linear/MLP evaluation regimes are distinct; threshold selection uses folds. Individual claims	Cell-DINO: Self-supervised image-based embeddings for cell fluorescent microscopy Original source ↗ Methods: Training and evaluation protocol on HPA datasets; Kaggle evaluation protocol; cached text lines 46, 86–94; task metric definitions and corresponding results table Version: version of record Retrieved: 2026-09-16T10:41:06Z	source checked automated source review · 2026-09-16 Audit details Targeted full-paper and supplement review of the outstanding task fields, with original dataset metadata checked where accessible. Source-scoped omissions are explicit; no independent benchmark reproduction or numerical-result change. Field: `attributes.profile.facts.10.value` Source artifact SHA-256: `12a53a78c70b3033c3351cf7afd4da42ebc98bb3281308f07e71e5baffc153a0` Hash scope: Hash scope not separately documented; inspect source record Inspected artifact
Metrics F1 for protein-localization classification; cell-line accuracy and Cell Painting metrics are separate tasks. Individual claims	Cell-DINO: Self-supervised image-based embeddings for cell fluorescent microscopy Original source ↗ Methods: Training and evaluation protocol on HPA datasets; Kaggle evaluation protocol; cached text lines 46, 86–94; task metric definitions and corresponding results table Version: version of record Retrieved: 2026-09-16T10:41:06Z	source checked automated source review · 2026-09-16 Audit details Targeted full-paper and supplement review of the outstanding task fields, with original dataset metadata checked where accessible. Source-scoped omissions are explicit; no independent benchmark reproduction or numerical-result change. Field: `attributes.profile.facts.2.value` Source artifact SHA-256: `12a53a78c70b3033c3351cf7afd4da42ebc98bb3281308f07e71e5baffc153a0` Hash scope: Hash scope not separately documented; inspect source record Inspected artifact
Baselines The paper includes nearest-neighbour, linear and MLP-based evaluations; these are distinct classifier settings. Individual claims	Cell-DINO: Self-supervised image-based embeddings for cell fluorescent microscopy Original source ↗ Methods: Training and evaluation protocol on HPA datasets; Kaggle evaluation protocol; cached text lines 46, 86–94; task metric definitions and corresponding results table Version: version of record Retrieved: 2026-09-16T10:41:06Z	source checked automated source review · 2026-09-16 Audit details Targeted full-paper and supplement review of the outstanding task fields, with original dataset metadata checked where accessible. Source-scoped omissions are explicit; no independent benchmark reproduction or numerical-result change. Field: `attributes.profile.facts.3.value` Source artifact SHA-256: `12a53a78c70b3033c3351cf7afd4da42ebc98bb3281308f07e71e5baffc153a0` Hash scope: Hash scope not separately documented; inspect source record Inspected artifact
Leakage controls Routine HPA train/validation analysis and Kaggle public-test generalization are different settings; per-class thresholds are selected on validation folds rather than inferred from the public score. Individual claims	Cell-DINO: Self-supervised image-based embeddings for cell fluorescent microscopy Original source ↗ Methods: Training and evaluation protocol on HPA datasets; Kaggle evaluation protocol; cached text lines 46, 86–94; task metric definitions and corresponding results table Version: version of record Retrieved: 2026-09-16T10:41:06Z	source checked automated source review · 2026-09-16 Audit details Targeted full-paper and supplement review of the outstanding task fields, with original dataset metadata checked where accessible. Source-scoped omissions are explicit; no independent benchmark reproduction or numerical-result change. Field: `attributes.profile.facts.4.value` Source artifact SHA-256: `12a53a78c70b3033c3351cf7afd4da42ebc98bb3281308f07e71e5baffc153a0` Hash scope: Hash scope not separately documented; inspect source record Inspected artifact
Uncertainty The protein-localization tables report point F1 or challenge scores. The HPA evaluation methods and S1 Text do not define repeated-seed confidence intervals or an uncertainty estimator for these values; comparisons between protocols retain their distinct classifiers and label access. Individual claims	Cell-DINO: Self-supervised image-based embeddings for cell fluorescent microscopy Original source ↗ Methods: Training and evaluation protocol on HPA datasets; Tables 1–3 and 6; S1 Text Shared locator for this statement’s cited sources; not a separate locator for each citation. Version: version of record Retrieved: 2026-09-16T10:41:06Z	unreported automated source review · 2026-09-16 Audit details Targeted full-paper and supplement review of the outstanding task fields, with original dataset metadata checked where accessible. Source-scoped omissions are explicit; no independent benchmark reproduction or numerical-result change. Field: `attributes.profile.facts.5.value` Source artifact SHA-256: `12a53a78c70b3033c3351cf7afd4da42ebc98bb3281308f07e71e5baffc153a0` Hash scope: Hash scope not separately documented; inspect source record Inspected artifact

All evidence (CSV)All evidence (JSONL)How to read the evidence

Sources and history

Release 2026-09-17-d277315f7d76 · Record review: needs review

3 source records and release history

Cell-DINO: Self-supervised image-based embeddings for cell fluorescent microscopy · Original source · version of record
Cell-DINO: Self-supervised image-based embeddings for cell fluorescent microscopy · Original source · version of record
cell-dino-2025__pcbi.1013828.s001.pdf · Original source · Retrieved 2026-09-16; sha256:e19d3d5c8d9dc1abcaf699c571349d869a1a7fa8e293660f49f67819f4b4d283

Download this release

Technical metadata and extraction receipts

Stable ID: reported-task-7621fa1be55362

areas: cells-tissues
tasks: protein localization classification
entity level: task
version: Not reported
task: protein localization classification
scope note: Paper-specific evaluation task; protocol completeness requires further extraction.
benchmark research: review date: 2026-09-17; status: primary_comparison_table_screened; primary sources: expansion-p3-cell-dino-2025; inspected locators: Tables 2 and 6; Methods: Kaggle evaluation protocol; searched queries: "PMC12826486"; gaps: Table 2 architectures and pretraining datasets differ; HPA-FoV and HPA-SC are separate protocols.; Do not attach Cell Painting MoA AUPRC to protein localization.; Table 6 ensembling is five MLPs with threshold/prediction averaging; not the base frozen model alone.; claim scope: Primary-paper discovery and source inspection. Source-checked results are not independently reproduced experiments.
historical missing metadata: protocol version: not_reported_in_legacy_extract; split: not_reported_in_legacy_extract
metadata review scope: historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.
legacy kinds: benchmark
entity classification: review date: 2026-09-17; rationale: This source-scoped record identifies the biological prediction task and holds its paper context. Preserve the existing task identity; exact split, model adaptation and scoring remain in linked evaluations or separate protocol records.; source ids: cell-dino-2025; source locator: Methods: Training and evaluation protocol on HPA datasets; Kaggle evaluation protocol; cached text lines 46, 86–94; task metric definitions and corresponding results table; ambiguities: A paper- or suite-specific task may constrain some inputs or metrics; that alone does not make it interchangeable with a complete versioned protocol. No protocol equivalence is inferred.; Some legacy profile Entity type facts use the generic phrase computational evaluation protocol. That boilerplate is not sufficient to establish a single fixed protocol identity or to merge this task with another protocol record.

Related records

dataset: HPA-FoV
benchmark: Cell-DINO ViT-L: protein localization classification

Suggest a correction