Protocol

Replogle RPE1: expression prediction for unseen single perturbations (Ahlmann-Eltze et al. 2025 Fig. 2a and Extended Data Fig. 8a)

Models are trained on the perturbations in the training part of a GEARS 'simulation' split and predict the mean expression profile for unseen single perturbations. Error is the L2 distance, and agreement the Pearson delta, between the mean predicted and observed expression over the 1,000 most highly expressed genes in the control condition, per perturbation.

7 evaluations · 14 results

Overview

An explanatory profile has not yet been reviewed for this record.

Consult the linked sources for architecture or protocol details. Missing evidence is not evidence of a missing capability.

7 recorded evaluations, 14 metric rows. A comparison chart has not yet been validated for these results. The table retains the individual findings and their sources.

View coverage and remaining gaps across all benchmarks

Results

Results are available, but no reviewed comparison panel is linked in this release.

All evaluations

7 evaluations · 14 results. Different protocols are not a single leaderboard.

Protocol: Replogle RPE1: expression prediction for unseen single perturbations (Ahlmann-Eltze et al. 2025 Fig. 2a and Extended Data Fig. 8a) · Dataset: Replogle et al. 2022 Perturb-seq essential-gene screen, RPE1 CRISPRi (GEARS distribution, Ahlmann-Eltze et al. 2025)

Sorted by L2 distance (1,000 most highly expressed control genes; mean over held-out perturbations and splits) (lower 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 configuration	L2 distance (1,000 most highly expressed control genes; mean over held-out perturbations and splits)	Pearson delta (1,000 most highly expressed control genes; mean over held-out perturbations and splits)
Linear model (G and P from training data) (Ahlmann-Eltze et al. 2025)Evaluation	5.39Computed by Rewire from the authors' source data (best in column)	0.68Computed by Rewire from the authors' source data (best in column)
Geneformer (Ahlmann-Eltze et al. 2025)Evaluation	5.49Computed by Rewire from the authors' source data	0.678Computed by Rewire from the authors' source data
UCE (Ahlmann-Eltze et al. 2025)Evaluation	5.51Computed by Rewire from the authors' source data	0.68Computed by Rewire from the authors' source data (best in column)
scBERT (Ahlmann-Eltze et al. 2025)Evaluation	5.53Computed by Rewire from the authors' source data	0.678Computed by Rewire from the authors' source data
Mean (Ahlmann-Eltze et al. 2025)Evaluation	5.53Computed by Rewire from the authors' source data	0.678Computed by Rewire from the authors' source data
scGPT (Ahlmann-Eltze et al. 2025)Evaluation	5.99Computed by Rewire from the authors' source data	0.588Computed by Rewire from the authors' source data
GEARS (Ahlmann-Eltze et al. 2025)Evaluation	6.11Computed by Rewire from the authors' source data	0.636Computed by Rewire from the authors' source data

All 14 result rows with coverage, uncertainty and sources

Filter evaluations

Tested model or methodProtocolDatasetEvaluation setupMetricEvidence origin

Applied filters: All linked evaluations

Exact evaluated configurations and original reported results
Tested configuration	Protocol and dataset	Finding	Evidence and details
Configuration: GEARS (Ahlmann-Eltze et al. 2025)	Protocol: Replogle RPE1: expression prediction for unseen single perturbations (Ahlmann-Eltze et al. 2025 Fig. 2a and Extended Data Fig. 8a) Dataset: Replogle et al. 2022 Perturb-seq essential-gene screen, RPE1 CRISPRi (GEARS distribution, Ahlmann-Eltze et al. 2025)	0.636 Pearson delta (1,000 most highly expressed control genes; mean over held-out perturbations and splits)Computed by Rewire from the authors' source data unitless · higher Uncertainty: Not reported by the source: No spread of the mean is stated; Rewire did not compute one Coverage: 413/413 perturbation-split rows scored	Independent external evaluation · Source checked Evidence concern: excluded from comparisons Panel A repeats Source Data Fig. 2 Panel A row for row (48, 262 and 413 rows per method over two splits). Methods restricts the analysis to 73, 398 and 629 perturbations predicted by all models; the sheet does not have these counts.Source Data Extended Data Fig. 8, Ahlmann-Eltze et al. 2025 · Methods 'Single perturbation benchmark setup' paragraph 4 versus Source Data Extended Data Fig. 8, sheet "Panel A" The Fig. 2 legend gives "134, 210 and 24 unseen single perturbations across two test-training splits". In Panel A these are the split 1 counts only (Replogle K562 134, Replogle RPE1 210, Adamson 24 rows per method); split 2 has 128, 203 and 24, so the two-split totals are 262, 413 and 48. The Fig. 2a means recorded here use all Panel A rows; whether the plotted red lines use both splits is not stated beyond the sheet itself. Reviewer check 2026-10-11: using split 1 rows only would change the per-model means by up to 0.61 (L2 distance) and 0.06 (Pearson delta), and would change the model with the lowest mean L2 distance on Adamson and Replogle RPE1, so the choice is material.Source Data Fig. 2, Ahlmann-Eltze et al. 2025 · Fig. 2 legend versus Source Data Fig. 2, sheet "Panel A" Methods restricts the single-perturbation analysis to perturbations predicted by all models: 73 for Adamson, 398 for Replogle K562 and 629 for Replogle RPE1. No sheet of Source Data Fig. 2 has these counts (Panel A: 48, 262 and 413 per method over two splits; Panel C: 20, 214 and 224 test or val rows per method).Source Data Fig. 2, Ahlmann-Eltze et al. 2025 · Methods 'Single perturbation benchmark setup' paragraph 4 versus Source Data Fig. 2, sheets "Panel A" and "Panel C" The test and val rows in Panel C (20, 214 and 224 per method for Adamson, Replogle K562 and Replogle RPE1) match neither the Fig. 2 legend counts nor the Methods counts. Fig. 2c was not extracted.Source Data Fig. 2, Ahlmann-Eltze et al. 2025 · Source Data Fig. 2, sheet "Panel C" versus the Fig. 2 legend and Methods The main text says CPA was not included in the single-perturbation benchmark, but method "cpa" has rows in Panel C. Fig. 2c was not extracted.Source Data Fig. 2, Ahlmann-Eltze et al. 2025 · Main text paragraph 16 versus Source Data Fig. 2, sheet "Panel C" Methods, coverage and source GEARS on Replogle RPE1 (Ahlmann-Eltze et al. 2025) perturbation-ahlmann-20261011-protocol-replogle-rpe1-single Aggregation: Not reported Deep-learning-based gene perturbation effect prediction does not yet outperform simple linear baselines; Source Data Extended Data Fig. 8, Ahlmann-Eltze et al. 2025 · Source Data Extended Data Fig. 8 (MOESM10), sheet "Panel A", rows with dataset_name "replogle_rpe1_essential" and method "gears", column "r2_delta"; plotted as the red line in Extended Data Fig. 8a
Configuration: Geneformer (Ahlmann-Eltze et al. 2025)	Protocol: Replogle RPE1: expression prediction for unseen single perturbations (Ahlmann-Eltze et al. 2025 Fig. 2a and Extended Data Fig. 8a) Dataset: Replogle et al. 2022 Perturb-seq essential-gene screen, RPE1 CRISPRi (GEARS distribution, Ahlmann-Eltze et al. 2025)	0.678 Pearson delta (1,000 most highly expressed control genes; mean over held-out perturbations and splits)Computed by Rewire from the authors' source data unitless · higher Uncertainty: Not reported by the source: No spread of the mean is stated; Rewire did not compute one Coverage: 413/413 perturbation-split rows scored	Independent external evaluation · Source checked Evidence concern: excluded from comparisons Panel A repeats Source Data Fig. 2 Panel A row for row (48, 262 and 413 rows per method over two splits). Methods restricts the analysis to 73, 398 and 629 perturbations predicted by all models; the sheet does not have these counts.Source Data Extended Data Fig. 8, Ahlmann-Eltze et al. 2025 · Methods 'Single perturbation benchmark setup' paragraph 4 versus Source Data Extended Data Fig. 8, sheet "Panel A" The Fig. 2 legend gives "134, 210 and 24 unseen single perturbations across two test-training splits". In Panel A these are the split 1 counts only (Replogle K562 134, Replogle RPE1 210, Adamson 24 rows per method); split 2 has 128, 203 and 24, so the two-split totals are 262, 413 and 48. The Fig. 2a means recorded here use all Panel A rows; whether the plotted red lines use both splits is not stated beyond the sheet itself. Reviewer check 2026-10-11: using split 1 rows only would change the per-model means by up to 0.61 (L2 distance) and 0.06 (Pearson delta), and would change the model with the lowest mean L2 distance on Adamson and Replogle RPE1, so the choice is material.Source Data Fig. 2, Ahlmann-Eltze et al. 2025 · Fig. 2 legend versus Source Data Fig. 2, sheet "Panel A" Methods restricts the single-perturbation analysis to perturbations predicted by all models: 73 for Adamson, 398 for Replogle K562 and 629 for Replogle RPE1. No sheet of Source Data Fig. 2 has these counts (Panel A: 48, 262 and 413 per method over two splits; Panel C: 20, 214 and 224 test or val rows per method).Source Data Fig. 2, Ahlmann-Eltze et al. 2025 · Methods 'Single perturbation benchmark setup' paragraph 4 versus Source Data Fig. 2, sheets "Panel A" and "Panel C" The test and val rows in Panel C (20, 214 and 224 per method for Adamson, Replogle K562 and Replogle RPE1) match neither the Fig. 2 legend counts nor the Methods counts. Fig. 2c was not extracted.Source Data Fig. 2, Ahlmann-Eltze et al. 2025 · Source Data Fig. 2, sheet "Panel C" versus the Fig. 2 legend and Methods The main text says CPA was not included in the single-perturbation benchmark, but method "cpa" has rows in Panel C. Fig. 2c was not extracted.Source Data Fig. 2, Ahlmann-Eltze et al. 2025 · Main text paragraph 16 versus Source Data Fig. 2, sheet "Panel C" Methods, coverage and source Geneformer on Replogle RPE1 (Ahlmann-Eltze et al. 2025) perturbation-ahlmann-20261011-protocol-replogle-rpe1-single Aggregation: Not reported Deep-learning-based gene perturbation effect prediction does not yet outperform simple linear baselines; Source Data Extended Data Fig. 8, Ahlmann-Eltze et al. 2025 · Source Data Extended Data Fig. 8 (MOESM10), sheet "Panel A", rows with dataset_name "replogle_rpe1_essential" and method "geneformer", column "r2_delta"; plotted as the red line in Extended Data Fig. 8a
Configuration: Linear model (G and P from training data) (Ahlmann-Eltze et al. 2025)	Protocol: Replogle RPE1: expression prediction for unseen single perturbations (Ahlmann-Eltze et al. 2025 Fig. 2a and Extended Data Fig. 8a) Dataset: Replogle et al. 2022 Perturb-seq essential-gene screen, RPE1 CRISPRi (GEARS distribution, Ahlmann-Eltze et al. 2025)	0.68 Pearson delta (1,000 most highly expressed control genes; mean over held-out perturbations and splits)Computed by Rewire from the authors' source data unitless · higher Uncertainty: Not reported by the source: No spread of the mean is stated; Rewire did not compute one Coverage: 413/413 perturbation-split rows scored	Author-reported evaluation · Source checked Evidence concern: excluded from comparisons Panel A repeats Source Data Fig. 2 Panel A row for row (48, 262 and 413 rows per method over two splits). Methods restricts the analysis to 73, 398 and 629 perturbations predicted by all models; the sheet does not have these counts.Source Data Extended Data Fig. 8, Ahlmann-Eltze et al. 2025 · Methods 'Single perturbation benchmark setup' paragraph 4 versus Source Data Extended Data Fig. 8, sheet "Panel A" The Fig. 2 legend gives "134, 210 and 24 unseen single perturbations across two test-training splits". In Panel A these are the split 1 counts only (Replogle K562 134, Replogle RPE1 210, Adamson 24 rows per method); split 2 has 128, 203 and 24, so the two-split totals are 262, 413 and 48. The Fig. 2a means recorded here use all Panel A rows; whether the plotted red lines use both splits is not stated beyond the sheet itself. Reviewer check 2026-10-11: using split 1 rows only would change the per-model means by up to 0.61 (L2 distance) and 0.06 (Pearson delta), and would change the model with the lowest mean L2 distance on Adamson and Replogle RPE1, so the choice is material.Source Data Fig. 2, Ahlmann-Eltze et al. 2025 · Fig. 2 legend versus Source Data Fig. 2, sheet "Panel A" Methods restricts the single-perturbation analysis to perturbations predicted by all models: 73 for Adamson, 398 for Replogle K562 and 629 for Replogle RPE1. No sheet of Source Data Fig. 2 has these counts (Panel A: 48, 262 and 413 per method over two splits; Panel C: 20, 214 and 224 test or val rows per method).Source Data Fig. 2, Ahlmann-Eltze et al. 2025 · Methods 'Single perturbation benchmark setup' paragraph 4 versus Source Data Fig. 2, sheets "Panel A" and "Panel C" The test and val rows in Panel C (20, 214 and 224 per method for Adamson, Replogle K562 and Replogle RPE1) match neither the Fig. 2 legend counts nor the Methods counts. Fig. 2c was not extracted.Source Data Fig. 2, Ahlmann-Eltze et al. 2025 · Source Data Fig. 2, sheet "Panel C" versus the Fig. 2 legend and Methods The main text says CPA was not included in the single-perturbation benchmark, but method "cpa" has rows in Panel C. Fig. 2c was not extracted.Source Data Fig. 2, Ahlmann-Eltze et al. 2025 · Main text paragraph 16 versus Source Data Fig. 2, sheet "Panel C" Methods, coverage and source Linear model (G and P from training data) on Replogle RPE1 (Ahlmann-Eltze et al. 2025) perturbation-ahlmann-20261011-protocol-replogle-rpe1-single Aggregation: Not reported Deep-learning-based gene perturbation effect prediction does not yet outperform simple linear baselines; Source Data Extended Data Fig. 8, Ahlmann-Eltze et al. 2025 · Source Data Extended Data Fig. 8 (MOESM10), sheet "Panel A", rows with dataset_name "replogle_rpe1_essential" and method "lpm_selftrained", column "r2_delta"; plotted as the red line in Extended Data Fig. 8a
Configuration: Mean (Ahlmann-Eltze et al. 2025)	Protocol: Replogle RPE1: expression prediction for unseen single perturbations (Ahlmann-Eltze et al. 2025 Fig. 2a and Extended Data Fig. 8a) Dataset: Replogle et al. 2022 Perturb-seq essential-gene screen, RPE1 CRISPRi (GEARS distribution, Ahlmann-Eltze et al. 2025)	0.678 Pearson delta (1,000 most highly expressed control genes; mean over held-out perturbations and splits)Computed by Rewire from the authors' source data unitless · higher Uncertainty: Not reported by the source: No spread of the mean is stated; Rewire did not compute one Coverage: 413/413 perturbation-split rows scored	Author-reported evaluation · Source checked Evidence concern: excluded from comparisons Panel A repeats Source Data Fig. 2 Panel A row for row (48, 262 and 413 rows per method over two splits). Methods restricts the analysis to 73, 398 and 629 perturbations predicted by all models; the sheet does not have these counts.Source Data Extended Data Fig. 8, Ahlmann-Eltze et al. 2025 · Methods 'Single perturbation benchmark setup' paragraph 4 versus Source Data Extended Data Fig. 8, sheet "Panel A" The Fig. 2 legend gives "134, 210 and 24 unseen single perturbations across two test-training splits". In Panel A these are the split 1 counts only (Replogle K562 134, Replogle RPE1 210, Adamson 24 rows per method); split 2 has 128, 203 and 24, so the two-split totals are 262, 413 and 48. The Fig. 2a means recorded here use all Panel A rows; whether the plotted red lines use both splits is not stated beyond the sheet itself. Reviewer check 2026-10-11: using split 1 rows only would change the per-model means by up to 0.61 (L2 distance) and 0.06 (Pearson delta), and would change the model with the lowest mean L2 distance on Adamson and Replogle RPE1, so the choice is material.Source Data Fig. 2, Ahlmann-Eltze et al. 2025 · Fig. 2 legend versus Source Data Fig. 2, sheet "Panel A" Methods restricts the single-perturbation analysis to perturbations predicted by all models: 73 for Adamson, 398 for Replogle K562 and 629 for Replogle RPE1. No sheet of Source Data Fig. 2 has these counts (Panel A: 48, 262 and 413 per method over two splits; Panel C: 20, 214 and 224 test or val rows per method).Source Data Fig. 2, Ahlmann-Eltze et al. 2025 · Methods 'Single perturbation benchmark setup' paragraph 4 versus Source Data Fig. 2, sheets "Panel A" and "Panel C" The test and val rows in Panel C (20, 214 and 224 per method for Adamson, Replogle K562 and Replogle RPE1) match neither the Fig. 2 legend counts nor the Methods counts. Fig. 2c was not extracted.Source Data Fig. 2, Ahlmann-Eltze et al. 2025 · Source Data Fig. 2, sheet "Panel C" versus the Fig. 2 legend and Methods The main text says CPA was not included in the single-perturbation benchmark, but method "cpa" has rows in Panel C. Fig. 2c was not extracted.Source Data Fig. 2, Ahlmann-Eltze et al. 2025 · Main text paragraph 16 versus Source Data Fig. 2, sheet "Panel C" Methods, coverage and source Mean on Replogle RPE1 (Ahlmann-Eltze et al. 2025) perturbation-ahlmann-20261011-protocol-replogle-rpe1-single Aggregation: Not reported Deep-learning-based gene perturbation effect prediction does not yet outperform simple linear baselines; Source Data Extended Data Fig. 8, Ahlmann-Eltze et al. 2025 · Source Data Extended Data Fig. 8 (MOESM10), sheet "Panel A", rows with dataset_name "replogle_rpe1_essential" and method "mean", column "r2_delta"; plotted as the red line in Extended Data Fig. 8a
Configuration: scBERT (Ahlmann-Eltze et al. 2025)	Protocol: Replogle RPE1: expression prediction for unseen single perturbations (Ahlmann-Eltze et al. 2025 Fig. 2a and Extended Data Fig. 8a) Dataset: Replogle et al. 2022 Perturb-seq essential-gene screen, RPE1 CRISPRi (GEARS distribution, Ahlmann-Eltze et al. 2025)	0.678 Pearson delta (1,000 most highly expressed control genes; mean over held-out perturbations and splits)Computed by Rewire from the authors' source data unitless · higher Uncertainty: Not reported by the source: No spread of the mean is stated; Rewire did not compute one Coverage: 413/413 perturbation-split rows scored	Independent external evaluation · Source checked Evidence concern: excluded from comparisons Panel A repeats Source Data Fig. 2 Panel A row for row (48, 262 and 413 rows per method over two splits). Methods restricts the analysis to 73, 398 and 629 perturbations predicted by all models; the sheet does not have these counts.Source Data Extended Data Fig. 8, Ahlmann-Eltze et al. 2025 · Methods 'Single perturbation benchmark setup' paragraph 4 versus Source Data Extended Data Fig. 8, sheet "Panel A" The Fig. 2 legend gives "134, 210 and 24 unseen single perturbations across two test-training splits". In Panel A these are the split 1 counts only (Replogle K562 134, Replogle RPE1 210, Adamson 24 rows per method); split 2 has 128, 203 and 24, so the two-split totals are 262, 413 and 48. The Fig. 2a means recorded here use all Panel A rows; whether the plotted red lines use both splits is not stated beyond the sheet itself. Reviewer check 2026-10-11: using split 1 rows only would change the per-model means by up to 0.61 (L2 distance) and 0.06 (Pearson delta), and would change the model with the lowest mean L2 distance on Adamson and Replogle RPE1, so the choice is material.Source Data Fig. 2, Ahlmann-Eltze et al. 2025 · Fig. 2 legend versus Source Data Fig. 2, sheet "Panel A" Methods restricts the single-perturbation analysis to perturbations predicted by all models: 73 for Adamson, 398 for Replogle K562 and 629 for Replogle RPE1. No sheet of Source Data Fig. 2 has these counts (Panel A: 48, 262 and 413 per method over two splits; Panel C: 20, 214 and 224 test or val rows per method).Source Data Fig. 2, Ahlmann-Eltze et al. 2025 · Methods 'Single perturbation benchmark setup' paragraph 4 versus Source Data Fig. 2, sheets "Panel A" and "Panel C" The test and val rows in Panel C (20, 214 and 224 per method for Adamson, Replogle K562 and Replogle RPE1) match neither the Fig. 2 legend counts nor the Methods counts. Fig. 2c was not extracted.Source Data Fig. 2, Ahlmann-Eltze et al. 2025 · Source Data Fig. 2, sheet "Panel C" versus the Fig. 2 legend and Methods The main text says CPA was not included in the single-perturbation benchmark, but method "cpa" has rows in Panel C. Fig. 2c was not extracted.Source Data Fig. 2, Ahlmann-Eltze et al. 2025 · Main text paragraph 16 versus Source Data Fig. 2, sheet "Panel C" Methods, coverage and source scBERT on Replogle RPE1 (Ahlmann-Eltze et al. 2025) perturbation-ahlmann-20261011-protocol-replogle-rpe1-single Aggregation: Not reported Deep-learning-based gene perturbation effect prediction does not yet outperform simple linear baselines; Source Data Extended Data Fig. 8, Ahlmann-Eltze et al. 2025 · Source Data Extended Data Fig. 8 (MOESM10), sheet "Panel A", rows with dataset_name "replogle_rpe1_essential" and method "scbert", column "r2_delta"; plotted as the red line in Extended Data Fig. 8a
Configuration: scGPT (Ahlmann-Eltze et al. 2025)	Protocol: Replogle RPE1: expression prediction for unseen single perturbations (Ahlmann-Eltze et al. 2025 Fig. 2a and Extended Data Fig. 8a) Dataset: Replogle et al. 2022 Perturb-seq essential-gene screen, RPE1 CRISPRi (GEARS distribution, Ahlmann-Eltze et al. 2025)	0.588 Pearson delta (1,000 most highly expressed control genes; mean over held-out perturbations and splits)Computed by Rewire from the authors' source data unitless · higher Uncertainty: Not reported by the source: No spread of the mean is stated; Rewire did not compute one Coverage: 413/413 perturbation-split rows scored	Independent external evaluation · Source checked Evidence concern: excluded from comparisons Panel A repeats Source Data Fig. 2 Panel A row for row (48, 262 and 413 rows per method over two splits). Methods restricts the analysis to 73, 398 and 629 perturbations predicted by all models; the sheet does not have these counts.Source Data Extended Data Fig. 8, Ahlmann-Eltze et al. 2025 · Methods 'Single perturbation benchmark setup' paragraph 4 versus Source Data Extended Data Fig. 8, sheet "Panel A" The Fig. 2 legend gives "134, 210 and 24 unseen single perturbations across two test-training splits". In Panel A these are the split 1 counts only (Replogle K562 134, Replogle RPE1 210, Adamson 24 rows per method); split 2 has 128, 203 and 24, so the two-split totals are 262, 413 and 48. The Fig. 2a means recorded here use all Panel A rows; whether the plotted red lines use both splits is not stated beyond the sheet itself. Reviewer check 2026-10-11: using split 1 rows only would change the per-model means by up to 0.61 (L2 distance) and 0.06 (Pearson delta), and would change the model with the lowest mean L2 distance on Adamson and Replogle RPE1, so the choice is material.Source Data Fig. 2, Ahlmann-Eltze et al. 2025 · Fig. 2 legend versus Source Data Fig. 2, sheet "Panel A" Methods restricts the single-perturbation analysis to perturbations predicted by all models: 73 for Adamson, 398 for Replogle K562 and 629 for Replogle RPE1. No sheet of Source Data Fig. 2 has these counts (Panel A: 48, 262 and 413 per method over two splits; Panel C: 20, 214 and 224 test or val rows per method).Source Data Fig. 2, Ahlmann-Eltze et al. 2025 · Methods 'Single perturbation benchmark setup' paragraph 4 versus Source Data Fig. 2, sheets "Panel A" and "Panel C" The test and val rows in Panel C (20, 214 and 224 per method for Adamson, Replogle K562 and Replogle RPE1) match neither the Fig. 2 legend counts nor the Methods counts. Fig. 2c was not extracted.Source Data Fig. 2, Ahlmann-Eltze et al. 2025 · Source Data Fig. 2, sheet "Panel C" versus the Fig. 2 legend and Methods The main text says CPA was not included in the single-perturbation benchmark, but method "cpa" has rows in Panel C. Fig. 2c was not extracted.Source Data Fig. 2, Ahlmann-Eltze et al. 2025 · Main text paragraph 16 versus Source Data Fig. 2, sheet "Panel C" Methods, coverage and source scGPT on Replogle RPE1 (Ahlmann-Eltze et al. 2025) perturbation-ahlmann-20261011-protocol-replogle-rpe1-single Aggregation: Not reported Deep-learning-based gene perturbation effect prediction does not yet outperform simple linear baselines; Source Data Extended Data Fig. 8, Ahlmann-Eltze et al. 2025 · Source Data Extended Data Fig. 8 (MOESM10), sheet "Panel A", rows with dataset_name "replogle_rpe1_essential" and method "scgpt", column "r2_delta"; plotted as the red line in Extended Data Fig. 8a
Configuration: UCE (Ahlmann-Eltze et al. 2025)	Protocol: Replogle RPE1: expression prediction for unseen single perturbations (Ahlmann-Eltze et al. 2025 Fig. 2a and Extended Data Fig. 8a) Dataset: Replogle et al. 2022 Perturb-seq essential-gene screen, RPE1 CRISPRi (GEARS distribution, Ahlmann-Eltze et al. 2025)	0.68 Pearson delta (1,000 most highly expressed control genes; mean over held-out perturbations and splits)Computed by Rewire from the authors' source data unitless · higher Uncertainty: Not reported by the source: No spread of the mean is stated; Rewire did not compute one Coverage: 413/413 perturbation-split rows scored	Independent external evaluation · Source checked Evidence concern: excluded from comparisons Panel A repeats Source Data Fig. 2 Panel A row for row (48, 262 and 413 rows per method over two splits). Methods restricts the analysis to 73, 398 and 629 perturbations predicted by all models; the sheet does not have these counts.Source Data Extended Data Fig. 8, Ahlmann-Eltze et al. 2025 · Methods 'Single perturbation benchmark setup' paragraph 4 versus Source Data Extended Data Fig. 8, sheet "Panel A" The Fig. 2 legend gives "134, 210 and 24 unseen single perturbations across two test-training splits". In Panel A these are the split 1 counts only (Replogle K562 134, Replogle RPE1 210, Adamson 24 rows per method); split 2 has 128, 203 and 24, so the two-split totals are 262, 413 and 48. The Fig. 2a means recorded here use all Panel A rows; whether the plotted red lines use both splits is not stated beyond the sheet itself. Reviewer check 2026-10-11: using split 1 rows only would change the per-model means by up to 0.61 (L2 distance) and 0.06 (Pearson delta), and would change the model with the lowest mean L2 distance on Adamson and Replogle RPE1, so the choice is material.Source Data Fig. 2, Ahlmann-Eltze et al. 2025 · Fig. 2 legend versus Source Data Fig. 2, sheet "Panel A" Methods restricts the single-perturbation analysis to perturbations predicted by all models: 73 for Adamson, 398 for Replogle K562 and 629 for Replogle RPE1. No sheet of Source Data Fig. 2 has these counts (Panel A: 48, 262 and 413 per method over two splits; Panel C: 20, 214 and 224 test or val rows per method).Source Data Fig. 2, Ahlmann-Eltze et al. 2025 · Methods 'Single perturbation benchmark setup' paragraph 4 versus Source Data Fig. 2, sheets "Panel A" and "Panel C" The test and val rows in Panel C (20, 214 and 224 per method for Adamson, Replogle K562 and Replogle RPE1) match neither the Fig. 2 legend counts nor the Methods counts. Fig. 2c was not extracted.Source Data Fig. 2, Ahlmann-Eltze et al. 2025 · Source Data Fig. 2, sheet "Panel C" versus the Fig. 2 legend and Methods The main text says CPA was not included in the single-perturbation benchmark, but method "cpa" has rows in Panel C. Fig. 2c was not extracted.Source Data Fig. 2, Ahlmann-Eltze et al. 2025 · Main text paragraph 16 versus Source Data Fig. 2, sheet "Panel C" Methods, coverage and source UCE on Replogle RPE1 (Ahlmann-Eltze et al. 2025) perturbation-ahlmann-20261011-protocol-replogle-rpe1-single Aggregation: Not reported Deep-learning-based gene perturbation effect prediction does not yet outperform simple linear baselines; Source Data Extended Data Fig. 8, Ahlmann-Eltze et al. 2025 · Source Data Extended Data Fig. 8 (MOESM10), sheet "Panel A", rows with dataset_name "replogle_rpe1_essential" and method "uce", column "r2_delta"; plotted as the red line in Extended Data Fig. 8a
Configuration: GEARS (Ahlmann-Eltze et al. 2025)	Protocol: Replogle RPE1: expression prediction for unseen single perturbations (Ahlmann-Eltze et al. 2025 Fig. 2a and Extended Data Fig. 8a) Dataset: Replogle et al. 2022 Perturb-seq essential-gene screen, RPE1 CRISPRi (GEARS distribution, Ahlmann-Eltze et al. 2025)	6.11 L2 distance (1,000 most highly expressed control genes; mean over held-out perturbations and splits)Computed by Rewire from the authors' source data unit-unreported · lower Uncertainty: Not reported by the source: No spread of the mean is stated; Rewire did not compute one Coverage: 413/413 perturbation-split rows scored	Independent external evaluation · Source checked Evidence concern: excluded from comparisons The Fig. 2 legend gives "134, 210 and 24 unseen single perturbations across two test-training splits". In Panel A these are the split 1 counts only (Replogle K562 134, Replogle RPE1 210, Adamson 24 rows per method); split 2 has 128, 203 and 24, so the two-split totals are 262, 413 and 48. The Fig. 2a means recorded here use all Panel A rows; whether the plotted red lines use both splits is not stated beyond the sheet itself. Reviewer check 2026-10-11: using split 1 rows only would change the per-model means by up to 0.61 (L2 distance) and 0.06 (Pearson delta), and would change the model with the lowest mean L2 distance on Adamson and Replogle RPE1, so the choice is material.Source Data Fig. 2, Ahlmann-Eltze et al. 2025 · Fig. 2 legend versus Source Data Fig. 2, sheet "Panel A" Methods restricts the single-perturbation analysis to perturbations predicted by all models: 73 for Adamson, 398 for Replogle K562 and 629 for Replogle RPE1. No sheet of Source Data Fig. 2 has these counts (Panel A: 48, 262 and 413 per method over two splits; Panel C: 20, 214 and 224 test or val rows per method).Source Data Fig. 2, Ahlmann-Eltze et al. 2025 · Methods 'Single perturbation benchmark setup' paragraph 4 versus Source Data Fig. 2, sheets "Panel A" and "Panel C" The test and val rows in Panel C (20, 214 and 224 per method for Adamson, Replogle K562 and Replogle RPE1) match neither the Fig. 2 legend counts nor the Methods counts. Fig. 2c was not extracted.Source Data Fig. 2, Ahlmann-Eltze et al. 2025 · Source Data Fig. 2, sheet "Panel C" versus the Fig. 2 legend and Methods The main text says CPA was not included in the single-perturbation benchmark, but method "cpa" has rows in Panel C. Fig. 2c was not extracted.Source Data Fig. 2, Ahlmann-Eltze et al. 2025 · Main text paragraph 16 versus Source Data Fig. 2, sheet "Panel C" Methods, coverage and source GEARS on Replogle RPE1 (Ahlmann-Eltze et al. 2025) perturbation-ahlmann-20261011-protocol-replogle-rpe1-single Aggregation: Not reported Deep-learning-based gene perturbation effect prediction does not yet outperform simple linear baselines; Source Data Fig. 2, Ahlmann-Eltze et al. 2025 · Source Data Fig. 2 (MOESM4), sheet "Panel A", rows with dataset_name "replogle_rpe1_essential" and method "gears", column "l2"; plotted as the red line in Fig. 2a
Configuration: Geneformer (Ahlmann-Eltze et al. 2025)	Protocol: Replogle RPE1: expression prediction for unseen single perturbations (Ahlmann-Eltze et al. 2025 Fig. 2a and Extended Data Fig. 8a) Dataset: Replogle et al. 2022 Perturb-seq essential-gene screen, RPE1 CRISPRi (GEARS distribution, Ahlmann-Eltze et al. 2025)	5.49 L2 distance (1,000 most highly expressed control genes; mean over held-out perturbations and splits)Computed by Rewire from the authors' source data unit-unreported · lower Uncertainty: Not reported by the source: No spread of the mean is stated; Rewire did not compute one Coverage: 413/413 perturbation-split rows scored	Independent external evaluation · Source checked Evidence concern: excluded from comparisons The Fig. 2 legend gives "134, 210 and 24 unseen single perturbations across two test-training splits". In Panel A these are the split 1 counts only (Replogle K562 134, Replogle RPE1 210, Adamson 24 rows per method); split 2 has 128, 203 and 24, so the two-split totals are 262, 413 and 48. The Fig. 2a means recorded here use all Panel A rows; whether the plotted red lines use both splits is not stated beyond the sheet itself. Reviewer check 2026-10-11: using split 1 rows only would change the per-model means by up to 0.61 (L2 distance) and 0.06 (Pearson delta), and would change the model with the lowest mean L2 distance on Adamson and Replogle RPE1, so the choice is material.Source Data Fig. 2, Ahlmann-Eltze et al. 2025 · Fig. 2 legend versus Source Data Fig. 2, sheet "Panel A" Methods restricts the single-perturbation analysis to perturbations predicted by all models: 73 for Adamson, 398 for Replogle K562 and 629 for Replogle RPE1. No sheet of Source Data Fig. 2 has these counts (Panel A: 48, 262 and 413 per method over two splits; Panel C: 20, 214 and 224 test or val rows per method).Source Data Fig. 2, Ahlmann-Eltze et al. 2025 · Methods 'Single perturbation benchmark setup' paragraph 4 versus Source Data Fig. 2, sheets "Panel A" and "Panel C" The test and val rows in Panel C (20, 214 and 224 per method for Adamson, Replogle K562 and Replogle RPE1) match neither the Fig. 2 legend counts nor the Methods counts. Fig. 2c was not extracted.Source Data Fig. 2, Ahlmann-Eltze et al. 2025 · Source Data Fig. 2, sheet "Panel C" versus the Fig. 2 legend and Methods The main text says CPA was not included in the single-perturbation benchmark, but method "cpa" has rows in Panel C. Fig. 2c was not extracted.Source Data Fig. 2, Ahlmann-Eltze et al. 2025 · Main text paragraph 16 versus Source Data Fig. 2, sheet "Panel C" Methods, coverage and source Geneformer on Replogle RPE1 (Ahlmann-Eltze et al. 2025) perturbation-ahlmann-20261011-protocol-replogle-rpe1-single Aggregation: Not reported Deep-learning-based gene perturbation effect prediction does not yet outperform simple linear baselines; Source Data Fig. 2, Ahlmann-Eltze et al. 2025 · Source Data Fig. 2 (MOESM4), sheet "Panel A", rows with dataset_name "replogle_rpe1_essential" and method "geneformer", column "l2"; plotted as the red line in Fig. 2a
Configuration: Linear model (G and P from training data) (Ahlmann-Eltze et al. 2025)	Protocol: Replogle RPE1: expression prediction for unseen single perturbations (Ahlmann-Eltze et al. 2025 Fig. 2a and Extended Data Fig. 8a) Dataset: Replogle et al. 2022 Perturb-seq essential-gene screen, RPE1 CRISPRi (GEARS distribution, Ahlmann-Eltze et al. 2025)	5.39 L2 distance (1,000 most highly expressed control genes; mean over held-out perturbations and splits)Computed by Rewire from the authors' source data unit-unreported · lower Uncertainty: Not reported by the source: No spread of the mean is stated; Rewire did not compute one Coverage: 413/413 perturbation-split rows scored	Author-reported evaluation · Source checked Evidence concern: excluded from comparisons The Fig. 2 legend gives "134, 210 and 24 unseen single perturbations across two test-training splits". In Panel A these are the split 1 counts only (Replogle K562 134, Replogle RPE1 210, Adamson 24 rows per method); split 2 has 128, 203 and 24, so the two-split totals are 262, 413 and 48. The Fig. 2a means recorded here use all Panel A rows; whether the plotted red lines use both splits is not stated beyond the sheet itself. Reviewer check 2026-10-11: using split 1 rows only would change the per-model means by up to 0.61 (L2 distance) and 0.06 (Pearson delta), and would change the model with the lowest mean L2 distance on Adamson and Replogle RPE1, so the choice is material.Source Data Fig. 2, Ahlmann-Eltze et al. 2025 · Fig. 2 legend versus Source Data Fig. 2, sheet "Panel A" Methods restricts the single-perturbation analysis to perturbations predicted by all models: 73 for Adamson, 398 for Replogle K562 and 629 for Replogle RPE1. No sheet of Source Data Fig. 2 has these counts (Panel A: 48, 262 and 413 per method over two splits; Panel C: 20, 214 and 224 test or val rows per method).Source Data Fig. 2, Ahlmann-Eltze et al. 2025 · Methods 'Single perturbation benchmark setup' paragraph 4 versus Source Data Fig. 2, sheets "Panel A" and "Panel C" The test and val rows in Panel C (20, 214 and 224 per method for Adamson, Replogle K562 and Replogle RPE1) match neither the Fig. 2 legend counts nor the Methods counts. Fig. 2c was not extracted.Source Data Fig. 2, Ahlmann-Eltze et al. 2025 · Source Data Fig. 2, sheet "Panel C" versus the Fig. 2 legend and Methods The main text says CPA was not included in the single-perturbation benchmark, but method "cpa" has rows in Panel C. Fig. 2c was not extracted.Source Data Fig. 2, Ahlmann-Eltze et al. 2025 · Main text paragraph 16 versus Source Data Fig. 2, sheet "Panel C" Methods, coverage and source Linear model (G and P from training data) on Replogle RPE1 (Ahlmann-Eltze et al. 2025) perturbation-ahlmann-20261011-protocol-replogle-rpe1-single Aggregation: Not reported Deep-learning-based gene perturbation effect prediction does not yet outperform simple linear baselines; Source Data Fig. 2, Ahlmann-Eltze et al. 2025 · Source Data Fig. 2 (MOESM4), sheet "Panel A", rows with dataset_name "replogle_rpe1_essential" and method "lpm_selftrained", column "l2"; plotted as the red line in Fig. 2a
Configuration: Mean (Ahlmann-Eltze et al. 2025)	Protocol: Replogle RPE1: expression prediction for unseen single perturbations (Ahlmann-Eltze et al. 2025 Fig. 2a and Extended Data Fig. 8a) Dataset: Replogle et al. 2022 Perturb-seq essential-gene screen, RPE1 CRISPRi (GEARS distribution, Ahlmann-Eltze et al. 2025)	5.53 L2 distance (1,000 most highly expressed control genes; mean over held-out perturbations and splits)Computed by Rewire from the authors' source data unit-unreported · lower Uncertainty: Not reported by the source: No spread of the mean is stated; Rewire did not compute one Coverage: 413/413 perturbation-split rows scored	Author-reported evaluation · Source checked Evidence concern: excluded from comparisons The Fig. 2 legend gives "134, 210 and 24 unseen single perturbations across two test-training splits". In Panel A these are the split 1 counts only (Replogle K562 134, Replogle RPE1 210, Adamson 24 rows per method); split 2 has 128, 203 and 24, so the two-split totals are 262, 413 and 48. The Fig. 2a means recorded here use all Panel A rows; whether the plotted red lines use both splits is not stated beyond the sheet itself. Reviewer check 2026-10-11: using split 1 rows only would change the per-model means by up to 0.61 (L2 distance) and 0.06 (Pearson delta), and would change the model with the lowest mean L2 distance on Adamson and Replogle RPE1, so the choice is material.Source Data Fig. 2, Ahlmann-Eltze et al. 2025 · Fig. 2 legend versus Source Data Fig. 2, sheet "Panel A" Methods restricts the single-perturbation analysis to perturbations predicted by all models: 73 for Adamson, 398 for Replogle K562 and 629 for Replogle RPE1. No sheet of Source Data Fig. 2 has these counts (Panel A: 48, 262 and 413 per method over two splits; Panel C: 20, 214 and 224 test or val rows per method).Source Data Fig. 2, Ahlmann-Eltze et al. 2025 · Methods 'Single perturbation benchmark setup' paragraph 4 versus Source Data Fig. 2, sheets "Panel A" and "Panel C" The test and val rows in Panel C (20, 214 and 224 per method for Adamson, Replogle K562 and Replogle RPE1) match neither the Fig. 2 legend counts nor the Methods counts. Fig. 2c was not extracted.Source Data Fig. 2, Ahlmann-Eltze et al. 2025 · Source Data Fig. 2, sheet "Panel C" versus the Fig. 2 legend and Methods The main text says CPA was not included in the single-perturbation benchmark, but method "cpa" has rows in Panel C. Fig. 2c was not extracted.Source Data Fig. 2, Ahlmann-Eltze et al. 2025 · Main text paragraph 16 versus Source Data Fig. 2, sheet "Panel C" Methods, coverage and source Mean on Replogle RPE1 (Ahlmann-Eltze et al. 2025) perturbation-ahlmann-20261011-protocol-replogle-rpe1-single Aggregation: Not reported Deep-learning-based gene perturbation effect prediction does not yet outperform simple linear baselines; Source Data Fig. 2, Ahlmann-Eltze et al. 2025 · Source Data Fig. 2 (MOESM4), sheet "Panel A", rows with dataset_name "replogle_rpe1_essential" and method "mean", column "l2"; plotted as the red line in Fig. 2a
Configuration: scBERT (Ahlmann-Eltze et al. 2025)	Protocol: Replogle RPE1: expression prediction for unseen single perturbations (Ahlmann-Eltze et al. 2025 Fig. 2a and Extended Data Fig. 8a) Dataset: Replogle et al. 2022 Perturb-seq essential-gene screen, RPE1 CRISPRi (GEARS distribution, Ahlmann-Eltze et al. 2025)	5.53 L2 distance (1,000 most highly expressed control genes; mean over held-out perturbations and splits)Computed by Rewire from the authors' source data unit-unreported · lower Uncertainty: Not reported by the source: No spread of the mean is stated; Rewire did not compute one Coverage: 413/413 perturbation-split rows scored	Independent external evaluation · Source checked Evidence concern: excluded from comparisons The Fig. 2 legend gives "134, 210 and 24 unseen single perturbations across two test-training splits". In Panel A these are the split 1 counts only (Replogle K562 134, Replogle RPE1 210, Adamson 24 rows per method); split 2 has 128, 203 and 24, so the two-split totals are 262, 413 and 48. The Fig. 2a means recorded here use all Panel A rows; whether the plotted red lines use both splits is not stated beyond the sheet itself. Reviewer check 2026-10-11: using split 1 rows only would change the per-model means by up to 0.61 (L2 distance) and 0.06 (Pearson delta), and would change the model with the lowest mean L2 distance on Adamson and Replogle RPE1, so the choice is material.Source Data Fig. 2, Ahlmann-Eltze et al. 2025 · Fig. 2 legend versus Source Data Fig. 2, sheet "Panel A" Methods restricts the single-perturbation analysis to perturbations predicted by all models: 73 for Adamson, 398 for Replogle K562 and 629 for Replogle RPE1. No sheet of Source Data Fig. 2 has these counts (Panel A: 48, 262 and 413 per method over two splits; Panel C: 20, 214 and 224 test or val rows per method).Source Data Fig. 2, Ahlmann-Eltze et al. 2025 · Methods 'Single perturbation benchmark setup' paragraph 4 versus Source Data Fig. 2, sheets "Panel A" and "Panel C" The test and val rows in Panel C (20, 214 and 224 per method for Adamson, Replogle K562 and Replogle RPE1) match neither the Fig. 2 legend counts nor the Methods counts. Fig. 2c was not extracted.Source Data Fig. 2, Ahlmann-Eltze et al. 2025 · Source Data Fig. 2, sheet "Panel C" versus the Fig. 2 legend and Methods The main text says CPA was not included in the single-perturbation benchmark, but method "cpa" has rows in Panel C. Fig. 2c was not extracted.Source Data Fig. 2, Ahlmann-Eltze et al. 2025 · Main text paragraph 16 versus Source Data Fig. 2, sheet "Panel C" Methods, coverage and source scBERT on Replogle RPE1 (Ahlmann-Eltze et al. 2025) perturbation-ahlmann-20261011-protocol-replogle-rpe1-single Aggregation: Not reported Deep-learning-based gene perturbation effect prediction does not yet outperform simple linear baselines; Source Data Fig. 2, Ahlmann-Eltze et al. 2025 · Source Data Fig. 2 (MOESM4), sheet "Panel A", rows with dataset_name "replogle_rpe1_essential" and method "scbert", column "l2"; plotted as the red line in Fig. 2a
Configuration: scGPT (Ahlmann-Eltze et al. 2025)	Protocol: Replogle RPE1: expression prediction for unseen single perturbations (Ahlmann-Eltze et al. 2025 Fig. 2a and Extended Data Fig. 8a) Dataset: Replogle et al. 2022 Perturb-seq essential-gene screen, RPE1 CRISPRi (GEARS distribution, Ahlmann-Eltze et al. 2025)	5.99 L2 distance (1,000 most highly expressed control genes; mean over held-out perturbations and splits)Computed by Rewire from the authors' source data unit-unreported · lower Uncertainty: Not reported by the source: No spread of the mean is stated; Rewire did not compute one Coverage: 413/413 perturbation-split rows scored	Independent external evaluation · Source checked Evidence concern: excluded from comparisons The Fig. 2 legend gives "134, 210 and 24 unseen single perturbations across two test-training splits". In Panel A these are the split 1 counts only (Replogle K562 134, Replogle RPE1 210, Adamson 24 rows per method); split 2 has 128, 203 and 24, so the two-split totals are 262, 413 and 48. The Fig. 2a means recorded here use all Panel A rows; whether the plotted red lines use both splits is not stated beyond the sheet itself. Reviewer check 2026-10-11: using split 1 rows only would change the per-model means by up to 0.61 (L2 distance) and 0.06 (Pearson delta), and would change the model with the lowest mean L2 distance on Adamson and Replogle RPE1, so the choice is material.Source Data Fig. 2, Ahlmann-Eltze et al. 2025 · Fig. 2 legend versus Source Data Fig. 2, sheet "Panel A" Methods restricts the single-perturbation analysis to perturbations predicted by all models: 73 for Adamson, 398 for Replogle K562 and 629 for Replogle RPE1. No sheet of Source Data Fig. 2 has these counts (Panel A: 48, 262 and 413 per method over two splits; Panel C: 20, 214 and 224 test or val rows per method).Source Data Fig. 2, Ahlmann-Eltze et al. 2025 · Methods 'Single perturbation benchmark setup' paragraph 4 versus Source Data Fig. 2, sheets "Panel A" and "Panel C" The test and val rows in Panel C (20, 214 and 224 per method for Adamson, Replogle K562 and Replogle RPE1) match neither the Fig. 2 legend counts nor the Methods counts. Fig. 2c was not extracted.Source Data Fig. 2, Ahlmann-Eltze et al. 2025 · Source Data Fig. 2, sheet "Panel C" versus the Fig. 2 legend and Methods The main text says CPA was not included in the single-perturbation benchmark, but method "cpa" has rows in Panel C. Fig. 2c was not extracted.Source Data Fig. 2, Ahlmann-Eltze et al. 2025 · Main text paragraph 16 versus Source Data Fig. 2, sheet "Panel C" Methods, coverage and source scGPT on Replogle RPE1 (Ahlmann-Eltze et al. 2025) perturbation-ahlmann-20261011-protocol-replogle-rpe1-single Aggregation: Not reported Deep-learning-based gene perturbation effect prediction does not yet outperform simple linear baselines; Source Data Fig. 2, Ahlmann-Eltze et al. 2025 · Source Data Fig. 2 (MOESM4), sheet "Panel A", rows with dataset_name "replogle_rpe1_essential" and method "scgpt", column "l2"; plotted as the red line in Fig. 2a
Configuration: UCE (Ahlmann-Eltze et al. 2025)	Protocol: Replogle RPE1: expression prediction for unseen single perturbations (Ahlmann-Eltze et al. 2025 Fig. 2a and Extended Data Fig. 8a) Dataset: Replogle et al. 2022 Perturb-seq essential-gene screen, RPE1 CRISPRi (GEARS distribution, Ahlmann-Eltze et al. 2025)	5.51 L2 distance (1,000 most highly expressed control genes; mean over held-out perturbations and splits)Computed by Rewire from the authors' source data unit-unreported · lower Uncertainty: Not reported by the source: No spread of the mean is stated; Rewire did not compute one Coverage: 413/413 perturbation-split rows scored	Independent external evaluation · Source checked Evidence concern: excluded from comparisons The Fig. 2 legend gives "134, 210 and 24 unseen single perturbations across two test-training splits". In Panel A these are the split 1 counts only (Replogle K562 134, Replogle RPE1 210, Adamson 24 rows per method); split 2 has 128, 203 and 24, so the two-split totals are 262, 413 and 48. The Fig. 2a means recorded here use all Panel A rows; whether the plotted red lines use both splits is not stated beyond the sheet itself. Reviewer check 2026-10-11: using split 1 rows only would change the per-model means by up to 0.61 (L2 distance) and 0.06 (Pearson delta), and would change the model with the lowest mean L2 distance on Adamson and Replogle RPE1, so the choice is material.Source Data Fig. 2, Ahlmann-Eltze et al. 2025 · Fig. 2 legend versus Source Data Fig. 2, sheet "Panel A" Methods restricts the single-perturbation analysis to perturbations predicted by all models: 73 for Adamson, 398 for Replogle K562 and 629 for Replogle RPE1. No sheet of Source Data Fig. 2 has these counts (Panel A: 48, 262 and 413 per method over two splits; Panel C: 20, 214 and 224 test or val rows per method).Source Data Fig. 2, Ahlmann-Eltze et al. 2025 · Methods 'Single perturbation benchmark setup' paragraph 4 versus Source Data Fig. 2, sheets "Panel A" and "Panel C" The test and val rows in Panel C (20, 214 and 224 per method for Adamson, Replogle K562 and Replogle RPE1) match neither the Fig. 2 legend counts nor the Methods counts. Fig. 2c was not extracted.Source Data Fig. 2, Ahlmann-Eltze et al. 2025 · Source Data Fig. 2, sheet "Panel C" versus the Fig. 2 legend and Methods The main text says CPA was not included in the single-perturbation benchmark, but method "cpa" has rows in Panel C. Fig. 2c was not extracted.Source Data Fig. 2, Ahlmann-Eltze et al. 2025 · Main text paragraph 16 versus Source Data Fig. 2, sheet "Panel C" Methods, coverage and source UCE on Replogle RPE1 (Ahlmann-Eltze et al. 2025) perturbation-ahlmann-20261011-protocol-replogle-rpe1-single Aggregation: Not reported Deep-learning-based gene perturbation effect prediction does not yet outperform simple linear baselines; Source Data Fig. 2, Ahlmann-Eltze et al. 2025 · Source Data Fig. 2 (MOESM4), sheet "Panel A", rows with dataset_name "replogle_rpe1_essential" and method "uce", column "l2"; plotted as the red line in Fig. 2a

Source checking is not independent reproduction. Release 2026-10-22-1b7cf027e4f6.

Methods and evaluation design

Procedure, tasks and evaluated configurations

Recorded evaluations

Each evaluation records what was tested and under which conditions.

Baseline coverage

Reference methods help show what a model adds beyond simple controls. We track a null control and a conventional method for each protocol.

0 of 2 active baseline roles have published Rewire measurements in this release. Measurements on a selected protocol do not establish coverage of an entire suite.

No execution recipe linked to this protocol. Recipe availability does not establish a completed evaluation.

Author-reported evaluations: 2
External evaluations: 5

Literature evidence is not a Rewire measurement. Executed but unpublished runs and private review status are not included.

Null control

Proposed control: requires review

No-change prediction under matched control conditions

Protocol-specific applicability, permitted inputs, access, split, evaluator and execution requirements need review before implementation or execution.

This is a suggested selection rule, not a validated method or a measured score.

Conventional reference

Proposed control: requires review

Training-only mean-effect or linear prediction

Protocol-specific applicability, permitted inputs, access, split, evaluator and execution requirements need review before implementation or execution.

This is a suggested selection rule, not a validated method or a measured score.

Protocol coverage CSV (gzip) · Model evaluation matrix (gzip) · Source table (gzip) · Release and checksums (gzip)

Coverage is derived from release 2026-10-22-1b7cf027e4f6. Source citations describe the original records; they do not validate an unreviewed baseline proposal. No results have been generated by this audit.

Run instructions

No runnable recipe has been reviewed for this protocol. Dataset access, model requirements, licences and compute requirements must be checked against its sources before execution.

Strengths, limitations and unresolved questions

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.

0 evidence rows matching the loaded filters

Claims, original sources and review scope · Release 2026-10-22-1b7cf027e4f6
Property and statement	Original source and location	Review and provenance

No evidence rows match these filters. Choose another scope or clear the search.

How to read the evidence

Sources and history

Release 2026-10-22-1b7cf027e4f6 · Record review: source checked

1 source record and release history

Deep-learning-based gene perturbation effect prediction does not yet outperform simple linear baselines · Original source · Nature Methods 22(8):1657, published online 2025-08-04; PMC12328236 full-text XML

Download this release (gzip)

Technical metadata and extraction receipts

Stable ID: perturbation-ahlmann-20261011-protocol-replogle-rpe1-single

areas: cells-tissues; single-cell
contexts: research
protocol: Models are trained on the perturbations in the training part of a GEARS 'simulation' split and predict the mean expression profile for unseen single perturbations. Error is the L2 distance, and agreement the Pearson delta, between the mean predicted and observed expression over the 1,000 most highly expressed genes in the control condition, per perturbation.
scope note: Split: GEARS 'simulation' test-training split, run twice (two splits). 413 held-out perturbation-split rows per model in the Source Data.
denominator: 413
metric definition: L2 distance: square root of the summed squared differences over the 1,000 most highly expressed genes in the control condition. Pearson delta: correlation of predicted and observed expression after subtracting the control expression (Methods 'Evaluation metrics').
aggregation: Mean per model over held-out perturbations, pooling the rows labelled test and val and all splits ("The horizontal red lines show the mean per model").
source locator: Main text paragraphs 13 and 16; Fig. 2 legend; Extended Data Fig. 8 legend; Methods 'Single perturbation benchmark setup' and 'Evaluation metrics'
limitations: The paper prints no per-model value; the means are computed by Rewire from the authors' Source Data rows and have not been checked against a printed number.; Four datasets, all from cancer cell lines, chosen because the GEARS, scGPT and scFoundation publications used them; the original quality control was not improved, so perturbations that did not affect their own target gene are included (main text paragraph 22).; The authors ran every model themselves with default parameters as far as possible and asked the model developers to review their code (Methods 'Software versions and parameters').; The Fig. 2 legend's perturbation counts are the split 1 counts only, and the Methods counts of perturbations predicted by all models (73, 398 and 629) match no Source Data sheet; see the evidence concerns on the Source Data Fig. 2 and Extended Data Fig. 8 source records.; scFoundation and CPA are not in this benchmark (main text paragraph 16).; UCE and scBERT post-perturbation embeddings use ground-truth expression of the perturbed genes; the authors note for UCE that this could leak test data (Methods 'Software versions and parameters').
missing metadata: uncertainty: reason: unreported; note: The figures show per-perturbation points and the mean; no spread of the mean is stated

Related records

uses data: Replogle et al. 2022 Perturb-seq essential-gene screen, RPE1 CRISPRi (GEARS distribution, Ahlmann-Eltze et al. 2025)
subject: reported_finding: perturbation-ahlmann-20261011-protocol-replogle-rpe1-single
assessment: GEARS on Replogle RPE1 (Ahlmann-Eltze et al. 2025)
assessment: Geneformer on Replogle RPE1 (Ahlmann-Eltze et al. 2025)
assessment: Linear model (G and P from training data) on Replogle RPE1 (Ahlmann-Eltze et al. 2025)
assessment: Mean on Replogle RPE1 (Ahlmann-Eltze et al. 2025)
assessment: scBERT on Replogle RPE1 (Ahlmann-Eltze et al. 2025)
assessment: scGPT on Replogle RPE1 (Ahlmann-Eltze et al. 2025)
assessment: UCE on Replogle RPE1 (Ahlmann-Eltze et al. 2025)
assessed by: Assess models for genetic perturbation experiments

Suggest a correction