rewire.itbenchmarks
Configuration

CPA* (scGPT)

Published model reimplemented or adapted by the PerturBench authors, as their asterisk marks.

8 evaluations · 8 results

Overview

Published model reimplemented or adapted by the PerturBench authors, as their asterisk marks.

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

Evaluations and results

8 evaluations · 8 results. Different protocols are not a single leaderboard.

Filter evaluations

Applied filters: All linked evaluations

Exact evaluated configurations and original reported results
Tested configurationProtocol and datasetFindingEvidence and details
Configuration: CPA* (scGPT)Task: PerturBench CB-COSINE: combination prediction on Norman19, Cosine similarity of log fold change
Dataset subset: Norman19 (PerturBench split)
0.70 ± 1 × 10 − 2 cosine_logfc
fraction · higher

Uncertainty: type: standard_deviation; value: 0.01

Coverage: Not reported scored / Not reported eligible

Author-reported evaluation · Source checked
Methods, coverage and source

CPA* (scGPT) on PerturBench CB-COSINE: combination prediction on Norman19, Cosine similarity of log fold change

Predict the effect of a pair of gene overexpressions from single perturbations, reported as the mean and one standard deviation over seeds.

Aggregation: Not reported

PerturBench: Benchmarking Machine Learning Models for Cellular Perturbation Analysis · Table 3, row(CPA ∗ (scGPT)), column(Cosine, log fold change (LogFC))
Configuration: CPA* (scGPT)Task: PerturBench CB-COSINE-RANK: combination prediction on Norman19, Cosine LogFC rank
Dataset subset: Norman19 (PerturBench split)
0.064 ± 1 × 10 − 2 cosine_logfc_rank
fraction · lower

Uncertainty: type: standard_deviation; value: 0.01

Coverage: Not reported scored / Not reported eligible

Author-reported evaluation · Source checked
Methods, coverage and source

CPA* (scGPT) on PerturBench CB-COSINE-RANK: combination prediction on Norman19, Cosine LogFC rank

Predict the effect of a pair of gene overexpressions from single perturbations, reported as the mean and one standard deviation over seeds.

Aggregation: Not reported

PerturBench: Benchmarking Machine Learning Models for Cellular Perturbation Analysis · Table 3, row(CPA ∗ (scGPT)), column(Cosine, LogFC rank)
Configuration: CPA* (scGPT)Task: PerturBench CB-RMSE-RANK: combination prediction on Norman19, RMSE mean rank
Dataset subset: Norman19 (PerturBench split)
0.13 ± 2 × 10 − 2 rmse_mean_rank
fraction · lower

Uncertainty: type: standard_deviation; value: 0.02

Coverage: Not reported scored / Not reported eligible

Author-reported evaluation · Source checked
Methods, coverage and source

CPA* (scGPT) on PerturBench CB-RMSE-RANK: combination prediction on Norman19, RMSE mean rank

Predict the effect of a pair of gene overexpressions from single perturbations, reported as the mean and one standard deviation over seeds.

Aggregation: Not reported

PerturBench: Benchmarking Machine Learning Models for Cellular Perturbation Analysis · Table 3, row(CPA ∗ (scGPT)), column(RMSE, mean rank)
Configuration: CPA* (scGPT)Task: PerturBench CB-RMSE: combination prediction on Norman19, RMSE of the mean
Dataset subset: Norman19 (PerturBench split)
0.061 ± 2 × 10 − 3 rmse_mean
error · lower

Uncertainty: type: standard_deviation; value: 0.002

Coverage: Not reported scored / Not reported eligible

Author-reported evaluation · Source checked
Methods, coverage and source

CPA* (scGPT) on PerturBench CB-RMSE: combination prediction on Norman19, RMSE of the mean

Predict the effect of a pair of gene overexpressions from single perturbations, reported as the mean and one standard deviation over seeds.

Aggregation: Not reported

PerturBench: Benchmarking Machine Learning Models for Cellular Perturbation Analysis · Table 3, row(CPA ∗ (scGPT)), column(RMSE, mean)
Configuration: CPA* (scGPT)Task: PerturBench CT-COSINE: covariate transfer on Srivatsan20, Cosine similarity of log fold change
Dataset subset: Srivatsan20 (PerturBench split)
0.29 ± 9 × 10 − 4 cosine_logfc
fraction · higher

Uncertainty: type: standard_deviation; value: 0.0009

Coverage: Not reported scored / Not reported eligible

Author-reported evaluation · Source checked
Methods, coverage and source

CPA* (scGPT) on PerturBench CT-COSINE: covariate transfer on Srivatsan20, Cosine similarity of log fold change

Train on some cell types and predict drug effects in a held-out cell type, reported as the mean and one standard deviation over seeds.

Aggregation: Not reported

PerturBench: Benchmarking Machine Learning Models for Cellular Perturbation Analysis · Table 2, row(CPA ∗ (scGPT)), column(Cosine, log fold change (LogFC))
Configuration: CPA* (scGPT)Task: PerturBench CT-COSINE-RANK: covariate transfer on Srivatsan20, Cosine LogFC rank
Dataset subset: Srivatsan20 (PerturBench split)
0.38 ± 2 × 10 − 2 cosine_logfc_rank
fraction · lower

Uncertainty: type: standard_deviation; value: 0.02

Coverage: Not reported scored / Not reported eligible

Author-reported evaluation · Source checked
Methods, coverage and source

CPA* (scGPT) on PerturBench CT-COSINE-RANK: covariate transfer on Srivatsan20, Cosine LogFC rank

Train on some cell types and predict drug effects in a held-out cell type, reported as the mean and one standard deviation over seeds.

Aggregation: Not reported

PerturBench: Benchmarking Machine Learning Models for Cellular Perturbation Analysis · Table 2, row(CPA ∗ (scGPT)), column(Cosine, LogFC rank)
Configuration: CPA* (scGPT)Task: PerturBench CT-RMSE-RANK: covariate transfer on Srivatsan20, RMSE mean rank
Dataset subset: Srivatsan20 (PerturBench split)
0.32 ± 1 × 10 − 2 rmse_mean_rank
fraction · lower

Uncertainty: type: standard_deviation; value: 0.01

Coverage: Not reported scored / Not reported eligible

Author-reported evaluation · Source checked
Methods, coverage and source

CPA* (scGPT) on PerturBench CT-RMSE-RANK: covariate transfer on Srivatsan20, RMSE mean rank

Train on some cell types and predict drug effects in a held-out cell type, reported as the mean and one standard deviation over seeds.

Aggregation: Not reported

PerturBench: Benchmarking Machine Learning Models for Cellular Perturbation Analysis · Table 2, row(CPA ∗ (scGPT)), column(RMSE, mean rank)
Configuration: CPA* (scGPT)Task: PerturBench CT-RMSE: covariate transfer on Srivatsan20, RMSE of the mean
Dataset subset: Srivatsan20 (PerturBench split)
0.021 ± 3 × 10 − 4 rmse_mean
error · lower

Uncertainty: type: standard_deviation; value: 0.0003

Coverage: Not reported scored / Not reported eligible

Author-reported evaluation · Source checked
Methods, coverage and source

CPA* (scGPT) on PerturBench CT-RMSE: covariate transfer on Srivatsan20, RMSE of the mean

Train on some cell types and predict drug effects in a held-out cell type, reported as the mean and one standard deviation over seeds.

Aggregation: Not reported

PerturBench: Benchmarking Machine Learning Models for Cellular Perturbation Analysis · Table 2, row(CPA ∗ (scGPT)), column(RMSE, mean)

Source checking is not independent reproduction. Release 2026-09-29-06401fd5b220.

Use this model

How it works, versions and access

Underlying model: scGPT. Results on this page belong to this configuration and its evaluated settings.

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.

1 evidence row matching the loaded filters

Claims, original sources and review scope · Release 2026-09-29-06401fd5b220
Property and statementOriginal source and locationReview and provenance
Relationship: uses model
catalog-model-scgpt
Individual claims
PerturBench: Benchmarking Machine Learning Models for Cellular Perturbation Analysis

Original source ↗

Section 3.2 Models, paragraph S3.SS2.p1.1 and Appendix E.5 scGPT Embeddings

Version: Primary full-text snapshot retrieved 2026-09-17; exact bytes pinned by SHA-256
Retrieved: 2026-09-17T08:06:28.400469+00:00

source checked

automated source review · 2026-09-23

Audit details

Source review establishes this relationship only. Exact evaluated configurations and original numerical review status remain unchanged. CPA* is fed scGPT embeddings; this is a downstream perturbation pipeline, not a scGPT checkpoint.

Field: links:uses_model:catalog-model-scgpt

Claim: model-evaluation-identity-801d4c7c01690df0754b

Source artifact SHA-256: 5c4804565dd9faa17a11853a79e9847dcb6da73715b4c91f62e5b274cc79f186

Hash scope: Exact retrieved primary paper artifact bytes.

Inspected artifact

Sources and history

View linked audit checks and correction history

Release 2026-09-29-06401fd5b220 · Record review: source checked

1 source records and release historyDownload this release
Technical metadata and extraction receipts

Stable ID: perturbench-method-cpa-scgpt

areas
cells-tissues
source locator
Table 2, row(CPA ∗ (scGPT))
missing metadata
checkpoint revision: unreported; parameters: unextracted
Related records

Suggest a correction