rewirebio.iobenchmarks
← Browse all protocols
Protocol

Combination phenotype prediction from small training sets: test RMSE at fixed average gene frequency

Models are trained on a subsample of measured gene pairs sized so that each gene is seen on average 4, 10 or 20 times, and scored by RMSE of predicted pair viability on a 30% held-out test set.

1 evaluation · 60 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.

1 recorded evaluation, 60 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

1 evaluation · 60 results. Different protocols are not a single leaderboard.

Protocol: Combination phenotype prediction from small training sets: test RMSE at fixed average gene frequency

Sorted by RMSE (test RMSE of pair viability, training set with each gene seen about 10 times on average) (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 configurationProtocol and datasetRMSE (test RMSE of pair viability, training set with each gene seen about 10 times on average)RMSE (test RMSE of pair viability, training set with each gene seen about 20 times on average)RMSE (test RMSE of pair viability, training set with each gene seen about 4 times on average)
NAIAD, trained once on a downsampled setEvaluationSimpson et al. 2023 combinatorial CRISPRi in K562, gene-level viability (as processed by Qin et al. 2025)2.0 (0.2) (best in column)1.9 (0.2) (best in column)2.2 (0.1) (best in column)
MLP on summed gene embeddings, trained once on a downsampled setEvaluationSimpson et al. 2023 combinatorial CRISPRi in K562, gene-level viability (as processed by Qin et al. 2025)2.4 (0.4)2.0 (0.3)4.3 (0.9)
RECOVER (gene adaptation), trained once on a downsampled setEvaluationSimpson et al. 2023 combinatorial CRISPRi in K562, gene-level viability (as processed by Qin et al. 2025)2.5 (0.2)1.9 (0.4) (best in column)3.9 (0.4)
NAIAD, trained once on a downsampled setEvaluationHorlbeck et al. 2018 combinatorial CRISPRi in Jurkat T cells, gene-level viability (as processed by Qin et al. 2025)3.0 (0.8)2.8 (0.6)3.0 (0.6)
Linear regression on single-gene effects, trained once on a downsampled setEvaluationSimpson et al. 2023 combinatorial CRISPRi in K562, gene-level viability (as processed by Qin et al. 2025)3.3 (0.2)3.3 (0.2)3.3 (0.3)
MLP on summed gene embeddings, trained once on a downsampled setEvaluationHorlbeck et al. 2018 combinatorial CRISPRi in Jurkat T cells, gene-level viability (as processed by Qin et al. 2025)3.5 (1.0)3.0 (0.4)5.5 (2.1)
RECOVER (gene adaptation), trained once on a downsampled setEvaluationHorlbeck et al. 2018 combinatorial CRISPRi in Jurkat T cells, gene-level viability (as processed by Qin et al. 2025)3.7 (1.0)3.0 (0.4)5.0 (0.7)
Linear regression on single-gene effects, trained once on a downsampled setEvaluationHorlbeck et al. 2018 combinatorial CRISPRi in Jurkat T cells, gene-level viability (as processed by Qin et al. 2025)3.8 (0.6)3.8 (0.6)3.9 (0.6)
NAIAD, trained once on a downsampled setEvaluationNorman et al. 2019 combinatorial CRISPRa in K562, gene-level viability (as processed by Qin et al. 2025)4.7 (1.2)4.7 (0.1)5.1 (1.8)
MLP on summed gene embeddings, trained once on a downsampled setEvaluationNorman et al. 2019 combinatorial CRISPRa in K562, gene-level viability (as processed by Qin et al. 2025)5.7 (1.3)5.0 (1.4)7.7 (0.9)
NAIAD, trained once on a downsampled setEvaluationHorlbeck et al. 2018 combinatorial CRISPRi in K562, gene-level viability (as processed by Qin et al. 2025)5.7 (0.8)5.4 (0.6)6.1 (1.9)
RECOVER (gene adaptation), trained once on a downsampled setEvaluationNorman et al. 2019 combinatorial CRISPRa in K562, gene-level viability (as processed by Qin et al. 2025)5.9 (1.0)4.7 (0.5)7.1 (2.8)
Linear regression on single-gene effects, trained once on a downsampled setEvaluationNorman et al. 2019 combinatorial CRISPRa in K562, gene-level viability (as processed by Qin et al. 2025)6.1 (1.3)6.1 (1.1)6.2 (1.7)
RECOVER (gene adaptation), trained once on a downsampled setEvaluationHorlbeck et al. 2018 combinatorial CRISPRi in K562, gene-level viability (as processed by Qin et al. 2025)6.1 (0.2)5.6 (1.0)7.9 (2.0)
MLP on summed gene embeddings, trained once on a downsampled setEvaluationHorlbeck et al. 2018 combinatorial CRISPRi in K562, gene-level viability (as processed by Qin et al. 2025)6.3 (1.1)5.9 (0.1)7.8 (2.5)
Linear regression on single-gene effects, trained once on a downsampled setEvaluationHorlbeck et al. 2018 combinatorial CRISPRi in K562, gene-level viability (as processed by Qin et al. 2025)6.4 (0.9)6.4 (0.9)6.4 (0.9)
GEARS, trained once on a downsampled setEvaluationSimpson et al. 2023 combinatorial CRISPRi in K562, gene-level viability (as processed by Qin et al. 2025)8.7 (10.3)3.5 (2.0)5.4 (3.4)
GEARS, trained once on a downsampled setEvaluationHorlbeck et al. 2018 combinatorial CRISPRi in K562, gene-level viability (as processed by Qin et al. 2025)12.0 (9.6)14.0 (14.0)13.0 (11.6)
GEARS, trained once on a downsampled setEvaluationNorman et al. 2019 combinatorial CRISPRa in K562, gene-level viability (as processed by Qin et al. 2025)15.5 (12.8)10.7 (12.1)16.6 (19.8)
GEARS, trained once on a downsampled setEvaluationHorlbeck et al. 2018 combinatorial CRISPRi in Jurkat T cells, gene-level viability (as processed by Qin et al. 2025)24.2 (28.5)20.7 (24.0)13.5 (16.1)
All 60 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
Configuration: GEARS, trained once on a downsampled setProtocol: Combination phenotype prediction from small training sets: test RMSE at fixed average gene frequency
Dataset: Horlbeck et al. 2018 combinatorial CRISPRi in Jurkat T cells, gene-level viability (as processed by Qin et al. 2025)		24.2 (28.5) RMSE (test RMSE of pair viability, training set with each gene seen about 10 times on average)
unitless · lower

Uncertainty: SE 28.5. Caption: error is SE across three cross-fold replicates

Coverage: Not reported scored / Not reported eligible

Independent external evaluation · Source checked
Methods, coverage and source

GEARS, trained once on a downsampled set on Horlbeck Jurkat (Tables 1 and 6, RMSE at fixed gene frequency)

pheno-20261022-qin2025-protocol-downsampling-rmse

Aggregation: Not reported

Qin et al. 2025, Active Learning for Efficient Discovery of Optimal Combinatorial Perturbations (ICML 2025, PMLR 267) · Table 6, Gene Frequency 10 block, row "GEARS", column "Horlbeck Jurkat"
Configuration: GEARS, trained once on a downsampled setProtocol: Combination phenotype prediction from small training sets: test RMSE at fixed average gene frequency
Dataset: Horlbeck et al. 2018 combinatorial CRISPRi in Jurkat T cells, gene-level viability (as processed by Qin et al. 2025)		20.7 (24.0) RMSE (test RMSE of pair viability, training set with each gene seen about 20 times on average)
unitless · lower

Uncertainty: SE 24.0. Caption: error is SE across three cross-fold replicates

Coverage: Not reported scored / Not reported eligible

Independent external evaluation · Source checked
Methods, coverage and source

GEARS, trained once on a downsampled set on Horlbeck Jurkat (Tables 1 and 6, RMSE at fixed gene frequency)

pheno-20261022-qin2025-protocol-downsampling-rmse

Aggregation: Not reported

Qin et al. 2025, Active Learning for Efficient Discovery of Optimal Combinatorial Perturbations (ICML 2025, PMLR 267) · Table 1, Gene Frequency 20 block, row "GEARS", column "Horlbeck Jurkat"
Configuration: GEARS, trained once on a downsampled setProtocol: Combination phenotype prediction from small training sets: test RMSE at fixed average gene frequency
Dataset: Horlbeck et al. 2018 combinatorial CRISPRi in Jurkat T cells, gene-level viability (as processed by Qin et al. 2025)		13.5 (16.1) RMSE (test RMSE of pair viability, training set with each gene seen about 4 times on average)
unitless · lower

Uncertainty: SE 16.1. Caption: error is SE across three cross-fold replicates

Coverage: Not reported scored / Not reported eligible

Independent external evaluation · Source checked
Methods, coverage and source

GEARS, trained once on a downsampled set on Horlbeck Jurkat (Tables 1 and 6, RMSE at fixed gene frequency)

pheno-20261022-qin2025-protocol-downsampling-rmse

Aggregation: Not reported

Qin et al. 2025, Active Learning for Efficient Discovery of Optimal Combinatorial Perturbations (ICML 2025, PMLR 267) · Table 1, Gene Frequency 4 block, row "GEARS", column "Horlbeck Jurkat"
Configuration: GEARS, trained once on a downsampled setProtocol: Combination phenotype prediction from small training sets: test RMSE at fixed average gene frequency
Dataset: Horlbeck et al. 2018 combinatorial CRISPRi in K562, gene-level viability (as processed by Qin et al. 2025)		12.0 (9.6) RMSE (test RMSE of pair viability, training set with each gene seen about 10 times on average)
unitless · lower

Uncertainty: SE 9.6. Caption: error is SE across three cross-fold replicates

Coverage: Not reported scored / Not reported eligible

Independent external evaluation · Source checked
Methods, coverage and source

GEARS, trained once on a downsampled set on Horlbeck K562 (Tables 1 and 6, RMSE at fixed gene frequency)

pheno-20261022-qin2025-protocol-downsampling-rmse

Aggregation: Not reported

Qin et al. 2025, Active Learning for Efficient Discovery of Optimal Combinatorial Perturbations (ICML 2025, PMLR 267) · Table 6, Gene Frequency 10 block, row "GEARS", column "Horlbeck K562"
Configuration: GEARS, trained once on a downsampled setProtocol: Combination phenotype prediction from small training sets: test RMSE at fixed average gene frequency
Dataset: Horlbeck et al. 2018 combinatorial CRISPRi in K562, gene-level viability (as processed by Qin et al. 2025)		14.0 (14.0) RMSE (test RMSE of pair viability, training set with each gene seen about 20 times on average)
unitless · lower

Uncertainty: SE 14.0. Caption: error is SE across three cross-fold replicates

Coverage: Not reported scored / Not reported eligible

Independent external evaluation · Source checked
Methods, coverage and source

GEARS, trained once on a downsampled set on Horlbeck K562 (Tables 1 and 6, RMSE at fixed gene frequency)

pheno-20261022-qin2025-protocol-downsampling-rmse

Aggregation: Not reported

Qin et al. 2025, Active Learning for Efficient Discovery of Optimal Combinatorial Perturbations (ICML 2025, PMLR 267) · Table 1, Gene Frequency 20 block, row "GEARS", column "Horlbeck K562"
Configuration: GEARS, trained once on a downsampled setProtocol: Combination phenotype prediction from small training sets: test RMSE at fixed average gene frequency
Dataset: Horlbeck et al. 2018 combinatorial CRISPRi in K562, gene-level viability (as processed by Qin et al. 2025)		13.0 (11.6) RMSE (test RMSE of pair viability, training set with each gene seen about 4 times on average)
unitless · lower

Uncertainty: SE 11.6. Caption: error is SE across three cross-fold replicates

Coverage: Not reported scored / Not reported eligible

Independent external evaluation · Source checked
Methods, coverage and source

GEARS, trained once on a downsampled set on Horlbeck K562 (Tables 1 and 6, RMSE at fixed gene frequency)

pheno-20261022-qin2025-protocol-downsampling-rmse

Aggregation: Not reported

Qin et al. 2025, Active Learning for Efficient Discovery of Optimal Combinatorial Perturbations (ICML 2025, PMLR 267) · Table 1, Gene Frequency 4 block, row "GEARS", column "Horlbeck K562"
Configuration: GEARS, trained once on a downsampled setProtocol: Combination phenotype prediction from small training sets: test RMSE at fixed average gene frequency
Dataset: Norman et al. 2019 combinatorial CRISPRa in K562, gene-level viability (as processed by Qin et al. 2025)		15.5 (12.8) RMSE (test RMSE of pair viability, training set with each gene seen about 10 times on average)
unitless · lower

Uncertainty: SE 12.8. Caption: error is SE across three cross-fold replicates

Coverage: Not reported scored / Not reported eligible

Independent external evaluation · Source checked
Methods, coverage and source

GEARS, trained once on a downsampled set on Norman (Tables 1 and 6, RMSE at fixed gene frequency)

pheno-20261022-qin2025-protocol-downsampling-rmse

Aggregation: Not reported

Qin et al. 2025, Active Learning for Efficient Discovery of Optimal Combinatorial Perturbations (ICML 2025, PMLR 267) · Table 6, Gene Frequency 10 block, row "GEARS", column "Norman"
Configuration: GEARS, trained once on a downsampled setProtocol: Combination phenotype prediction from small training sets: test RMSE at fixed average gene frequency
Dataset: Norman et al. 2019 combinatorial CRISPRa in K562, gene-level viability (as processed by Qin et al. 2025)		10.7 (12.1) RMSE (test RMSE of pair viability, training set with each gene seen about 20 times on average)
unitless · lower

Uncertainty: SE 12.1. Caption: error is SE across three cross-fold replicates

Coverage: Not reported scored / Not reported eligible

Independent external evaluation · Source checked
Methods, coverage and source

GEARS, trained once on a downsampled set on Norman (Tables 1 and 6, RMSE at fixed gene frequency)

pheno-20261022-qin2025-protocol-downsampling-rmse

Aggregation: Not reported

Qin et al. 2025, Active Learning for Efficient Discovery of Optimal Combinatorial Perturbations (ICML 2025, PMLR 267) · Table 1, Gene Frequency 20 block, row "GEARS", column "Norman"
Configuration: GEARS, trained once on a downsampled setProtocol: Combination phenotype prediction from small training sets: test RMSE at fixed average gene frequency
Dataset: Norman et al. 2019 combinatorial CRISPRa in K562, gene-level viability (as processed by Qin et al. 2025)		16.6 (19.8) RMSE (test RMSE of pair viability, training set with each gene seen about 4 times on average)
unitless · lower

Uncertainty: SE 19.8. Caption: error is SE across three cross-fold replicates

Coverage: Not reported scored / Not reported eligible

Independent external evaluation · Source checked
Methods, coverage and source

GEARS, trained once on a downsampled set on Norman (Tables 1 and 6, RMSE at fixed gene frequency)

pheno-20261022-qin2025-protocol-downsampling-rmse

Aggregation: Not reported

Qin et al. 2025, Active Learning for Efficient Discovery of Optimal Combinatorial Perturbations (ICML 2025, PMLR 267) · Table 1, Gene Frequency 4 block, row "GEARS", column "Norman"
Configuration: GEARS, trained once on a downsampled setProtocol: Combination phenotype prediction from small training sets: test RMSE at fixed average gene frequency
Dataset: Simpson et al. 2023 combinatorial CRISPRi in K562, gene-level viability (as processed by Qin et al. 2025)		8.7 (10.3) RMSE (test RMSE of pair viability, training set with each gene seen about 10 times on average)
unitless · lower

Uncertainty: SE 10.3. Caption: error is SE across three cross-fold replicates

Coverage: Not reported scored / Not reported eligible

Independent external evaluation · Source checked
Methods, coverage and source

GEARS, trained once on a downsampled set on Simpson (Tables 1 and 6, RMSE at fixed gene frequency)

pheno-20261022-qin2025-protocol-downsampling-rmse

Aggregation: Not reported

Qin et al. 2025, Active Learning for Efficient Discovery of Optimal Combinatorial Perturbations (ICML 2025, PMLR 267) · Table 6, Gene Frequency 10 block, row "GEARS", column "Simpson"
Configuration: GEARS, trained once on a downsampled setProtocol: Combination phenotype prediction from small training sets: test RMSE at fixed average gene frequency
Dataset: Simpson et al. 2023 combinatorial CRISPRi in K562, gene-level viability (as processed by Qin et al. 2025)		3.5 (2.0) RMSE (test RMSE of pair viability, training set with each gene seen about 20 times on average)
unitless · lower

Uncertainty: SE 2.0. Caption: error is SE across three cross-fold replicates

Coverage: Not reported scored / Not reported eligible

Independent external evaluation · Source checked
Methods, coverage and source

GEARS, trained once on a downsampled set on Simpson (Tables 1 and 6, RMSE at fixed gene frequency)

pheno-20261022-qin2025-protocol-downsampling-rmse

Aggregation: Not reported

Qin et al. 2025, Active Learning for Efficient Discovery of Optimal Combinatorial Perturbations (ICML 2025, PMLR 267) · Table 1, Gene Frequency 20 block, row "GEARS", column "Simpson"
Configuration: GEARS, trained once on a downsampled setProtocol: Combination phenotype prediction from small training sets: test RMSE at fixed average gene frequency
Dataset: Simpson et al. 2023 combinatorial CRISPRi in K562, gene-level viability (as processed by Qin et al. 2025)		5.4 (3.4) RMSE (test RMSE of pair viability, training set with each gene seen about 4 times on average)
unitless · lower

Uncertainty: SE 3.4. Caption: error is SE across three cross-fold replicates

Coverage: Not reported scored / Not reported eligible

Independent external evaluation · Source checked
Methods, coverage and source

GEARS, trained once on a downsampled set on Simpson (Tables 1 and 6, RMSE at fixed gene frequency)

pheno-20261022-qin2025-protocol-downsampling-rmse

Aggregation: Not reported

Qin et al. 2025, Active Learning for Efficient Discovery of Optimal Combinatorial Perturbations (ICML 2025, PMLR 267) · Table 1, Gene Frequency 4 block, row "GEARS", column "Simpson"
Configuration: Linear regression on single-gene effects, trained once on a downsampled setProtocol: Combination phenotype prediction from small training sets: test RMSE at fixed average gene frequency
Dataset: Horlbeck et al. 2018 combinatorial CRISPRi in Jurkat T cells, gene-level viability (as processed by Qin et al. 2025)		3.8 (0.6) RMSE (test RMSE of pair viability, training set with each gene seen about 10 times on average)
unitless · lower

Uncertainty: SE 0.6. Caption: error is SE across three cross-fold replicates

Coverage: Not reported scored / Not reported eligible

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

Linear regression on single-gene effects, trained once on a downsampled set on Horlbeck Jurkat (Tables 1 and 6, RMSE at fixed gene frequency)

pheno-20261022-qin2025-protocol-downsampling-rmse

Aggregation: Not reported

Qin et al. 2025, Active Learning for Efficient Discovery of Optimal Combinatorial Perturbations (ICML 2025, PMLR 267) · Table 6, Gene Frequency 10 block, row "Linear", column "Horlbeck Jurkat"
Configuration: Linear regression on single-gene effects, trained once on a downsampled setProtocol: Combination phenotype prediction from small training sets: test RMSE at fixed average gene frequency
Dataset: Horlbeck et al. 2018 combinatorial CRISPRi in Jurkat T cells, gene-level viability (as processed by Qin et al. 2025)		3.8 (0.6) RMSE (test RMSE of pair viability, training set with each gene seen about 20 times on average)
unitless · lower

Uncertainty: SE 0.6. Caption: error is SE across three cross-fold replicates

Coverage: Not reported scored / Not reported eligible

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

Linear regression on single-gene effects, trained once on a downsampled set on Horlbeck Jurkat (Tables 1 and 6, RMSE at fixed gene frequency)

pheno-20261022-qin2025-protocol-downsampling-rmse

Aggregation: Not reported

Qin et al. 2025, Active Learning for Efficient Discovery of Optimal Combinatorial Perturbations (ICML 2025, PMLR 267) · Table 1, Gene Frequency 20 block, row "Linear", column "Horlbeck Jurkat"
Configuration: Linear regression on single-gene effects, trained once on a downsampled setProtocol: Combination phenotype prediction from small training sets: test RMSE at fixed average gene frequency
Dataset: Horlbeck et al. 2018 combinatorial CRISPRi in Jurkat T cells, gene-level viability (as processed by Qin et al. 2025)		3.9 (0.6) RMSE (test RMSE of pair viability, training set with each gene seen about 4 times on average)
unitless · lower

Uncertainty: SE 0.6. Caption: error is SE across three cross-fold replicates

Coverage: Not reported scored / Not reported eligible

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

Linear regression on single-gene effects, trained once on a downsampled set on Horlbeck Jurkat (Tables 1 and 6, RMSE at fixed gene frequency)

pheno-20261022-qin2025-protocol-downsampling-rmse

Aggregation: Not reported

Qin et al. 2025, Active Learning for Efficient Discovery of Optimal Combinatorial Perturbations (ICML 2025, PMLR 267) · Table 1, Gene Frequency 4 block, row "Linear", column "Horlbeck Jurkat"
Configuration: Linear regression on single-gene effects, trained once on a downsampled setProtocol: Combination phenotype prediction from small training sets: test RMSE at fixed average gene frequency
Dataset: Horlbeck et al. 2018 combinatorial CRISPRi in K562, gene-level viability (as processed by Qin et al. 2025)		6.4 (0.9) RMSE (test RMSE of pair viability, training set with each gene seen about 10 times on average)
unitless · lower

Uncertainty: SE 0.9. Caption: error is SE across three cross-fold replicates

Coverage: Not reported scored / Not reported eligible

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

Linear regression on single-gene effects, trained once on a downsampled set on Horlbeck K562 (Tables 1 and 6, RMSE at fixed gene frequency)

pheno-20261022-qin2025-protocol-downsampling-rmse

Aggregation: Not reported

Qin et al. 2025, Active Learning for Efficient Discovery of Optimal Combinatorial Perturbations (ICML 2025, PMLR 267) · Table 6, Gene Frequency 10 block, row "Linear", column "Horlbeck K562"
Configuration: Linear regression on single-gene effects, trained once on a downsampled setProtocol: Combination phenotype prediction from small training sets: test RMSE at fixed average gene frequency
Dataset: Horlbeck et al. 2018 combinatorial CRISPRi in K562, gene-level viability (as processed by Qin et al. 2025)		6.4 (0.9) RMSE (test RMSE of pair viability, training set with each gene seen about 20 times on average)
unitless · lower

Uncertainty: SE 0.9. Caption: error is SE across three cross-fold replicates

Coverage: Not reported scored / Not reported eligible

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

Linear regression on single-gene effects, trained once on a downsampled set on Horlbeck K562 (Tables 1 and 6, RMSE at fixed gene frequency)

pheno-20261022-qin2025-protocol-downsampling-rmse

Aggregation: Not reported

Qin et al. 2025, Active Learning for Efficient Discovery of Optimal Combinatorial Perturbations (ICML 2025, PMLR 267) · Table 1, Gene Frequency 20 block, row "Linear", column "Horlbeck K562"
Configuration: Linear regression on single-gene effects, trained once on a downsampled setProtocol: Combination phenotype prediction from small training sets: test RMSE at fixed average gene frequency
Dataset: Horlbeck et al. 2018 combinatorial CRISPRi in K562, gene-level viability (as processed by Qin et al. 2025)		6.4 (0.9) RMSE (test RMSE of pair viability, training set with each gene seen about 4 times on average)
unitless · lower

Uncertainty: SE 0.9. Caption: error is SE across three cross-fold replicates

Coverage: Not reported scored / Not reported eligible

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

Linear regression on single-gene effects, trained once on a downsampled set on Horlbeck K562 (Tables 1 and 6, RMSE at fixed gene frequency)

pheno-20261022-qin2025-protocol-downsampling-rmse

Aggregation: Not reported

Qin et al. 2025, Active Learning for Efficient Discovery of Optimal Combinatorial Perturbations (ICML 2025, PMLR 267) · Table 1, Gene Frequency 4 block, row "Linear", column "Horlbeck K562"
Configuration: Linear regression on single-gene effects, trained once on a downsampled setProtocol: Combination phenotype prediction from small training sets: test RMSE at fixed average gene frequency
Dataset: Norman et al. 2019 combinatorial CRISPRa in K562, gene-level viability (as processed by Qin et al. 2025)		6.1 (1.3) RMSE (test RMSE of pair viability, training set with each gene seen about 10 times on average)
unitless · lower

Uncertainty: SE 1.3. Caption: error is SE across three cross-fold replicates

Coverage: Not reported scored / Not reported eligible

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

Linear regression on single-gene effects, trained once on a downsampled set on Norman (Tables 1 and 6, RMSE at fixed gene frequency)

pheno-20261022-qin2025-protocol-downsampling-rmse

Aggregation: Not reported

Qin et al. 2025, Active Learning for Efficient Discovery of Optimal Combinatorial Perturbations (ICML 2025, PMLR 267) · Table 6, Gene Frequency 10 block, row "Linear", column "Norman"
Configuration: Linear regression on single-gene effects, trained once on a downsampled setProtocol: Combination phenotype prediction from small training sets: test RMSE at fixed average gene frequency
Dataset: Norman et al. 2019 combinatorial CRISPRa in K562, gene-level viability (as processed by Qin et al. 2025)		6.1 (1.1) RMSE (test RMSE of pair viability, training set with each gene seen about 20 times on average)
unitless · lower

Uncertainty: SE 1.1. Caption: error is SE across three cross-fold replicates

Coverage: Not reported scored / Not reported eligible

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

Linear regression on single-gene effects, trained once on a downsampled set on Norman (Tables 1 and 6, RMSE at fixed gene frequency)

pheno-20261022-qin2025-protocol-downsampling-rmse

Aggregation: Not reported

Qin et al. 2025, Active Learning for Efficient Discovery of Optimal Combinatorial Perturbations (ICML 2025, PMLR 267) · Table 1, Gene Frequency 20 block, row "Linear", column "Norman"
Configuration: Linear regression on single-gene effects, trained once on a downsampled setProtocol: Combination phenotype prediction from small training sets: test RMSE at fixed average gene frequency
Dataset: Norman et al. 2019 combinatorial CRISPRa in K562, gene-level viability (as processed by Qin et al. 2025)		6.2 (1.7) RMSE (test RMSE of pair viability, training set with each gene seen about 4 times on average)
unitless · lower

Uncertainty: SE 1.7. Caption: error is SE across three cross-fold replicates

Coverage: Not reported scored / Not reported eligible

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

Linear regression on single-gene effects, trained once on a downsampled set on Norman (Tables 1 and 6, RMSE at fixed gene frequency)

pheno-20261022-qin2025-protocol-downsampling-rmse

Aggregation: Not reported

Qin et al. 2025, Active Learning for Efficient Discovery of Optimal Combinatorial Perturbations (ICML 2025, PMLR 267) · Table 1, Gene Frequency 4 block, row "Linear", column "Norman"
Configuration: Linear regression on single-gene effects, trained once on a downsampled setProtocol: Combination phenotype prediction from small training sets: test RMSE at fixed average gene frequency
Dataset: Simpson et al. 2023 combinatorial CRISPRi in K562, gene-level viability (as processed by Qin et al. 2025)		3.3 (0.2) RMSE (test RMSE of pair viability, training set with each gene seen about 10 times on average)
unitless · lower

Uncertainty: SE 0.2. Caption: error is SE across three cross-fold replicates

Coverage: Not reported scored / Not reported eligible

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

Linear regression on single-gene effects, trained once on a downsampled set on Simpson (Tables 1 and 6, RMSE at fixed gene frequency)

pheno-20261022-qin2025-protocol-downsampling-rmse

Aggregation: Not reported

Qin et al. 2025, Active Learning for Efficient Discovery of Optimal Combinatorial Perturbations (ICML 2025, PMLR 267) · Table 6, Gene Frequency 10 block, row "Linear", column "Simpson"
Configuration: Linear regression on single-gene effects, trained once on a downsampled setProtocol: Combination phenotype prediction from small training sets: test RMSE at fixed average gene frequency
Dataset: Simpson et al. 2023 combinatorial CRISPRi in K562, gene-level viability (as processed by Qin et al. 2025)		3.3 (0.2) RMSE (test RMSE of pair viability, training set with each gene seen about 20 times on average)
unitless · lower

Uncertainty: SE 0.2. Caption: error is SE across three cross-fold replicates

Coverage: Not reported scored / Not reported eligible

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

Linear regression on single-gene effects, trained once on a downsampled set on Simpson (Tables 1 and 6, RMSE at fixed gene frequency)

pheno-20261022-qin2025-protocol-downsampling-rmse

Aggregation: Not reported

Qin et al. 2025, Active Learning for Efficient Discovery of Optimal Combinatorial Perturbations (ICML 2025, PMLR 267) · Table 1, Gene Frequency 20 block, row "Linear", column "Simpson"
Configuration: Linear regression on single-gene effects, trained once on a downsampled setProtocol: Combination phenotype prediction from small training sets: test RMSE at fixed average gene frequency
Dataset: Simpson et al. 2023 combinatorial CRISPRi in K562, gene-level viability (as processed by Qin et al. 2025)		3.3 (0.3) RMSE (test RMSE of pair viability, training set with each gene seen about 4 times on average)
unitless · lower

Uncertainty: SE 0.3. Caption: error is SE across three cross-fold replicates

Coverage: Not reported scored / Not reported eligible

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

Linear regression on single-gene effects, trained once on a downsampled set on Simpson (Tables 1 and 6, RMSE at fixed gene frequency)

pheno-20261022-qin2025-protocol-downsampling-rmse

Aggregation: Not reported

Qin et al. 2025, Active Learning for Efficient Discovery of Optimal Combinatorial Perturbations (ICML 2025, PMLR 267) · Table 1, Gene Frequency 4 block, row "Linear", column "Simpson"
Configuration: MLP on summed gene embeddings, trained once on a downsampled setProtocol: Combination phenotype prediction from small training sets: test RMSE at fixed average gene frequency
Dataset: Horlbeck et al. 2018 combinatorial CRISPRi in Jurkat T cells, gene-level viability (as processed by Qin et al. 2025)		3.5 (1.0) RMSE (test RMSE of pair viability, training set with each gene seen about 10 times on average)
unitless · lower

Uncertainty: SE 1.0. Caption: error is SE across three cross-fold replicates

Coverage: Not reported scored / Not reported eligible

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

MLP on summed gene embeddings, trained once on a downsampled set on Horlbeck Jurkat (Tables 1 and 6, RMSE at fixed gene frequency)

pheno-20261022-qin2025-protocol-downsampling-rmse

Aggregation: Not reported

Qin et al. 2025, Active Learning for Efficient Discovery of Optimal Combinatorial Perturbations (ICML 2025, PMLR 267) · Table 6, Gene Frequency 10 block, row "MLP", column "Horlbeck Jurkat"

Source checking is not independent reproduction. Release 2026-10-23-5347695500ed.

Methods and evaluation design

Procedure, tasks and evaluated configurations

Recorded evaluations

Each evaluation records what was tested and under which conditions.

Explore all linked results

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
12
External evaluations
8

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

Null control

Proposed control: requires review

Select a task-valid null control after reviewing inputs and metric

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

Select an upstream conventional reference after reviewing the full protocol

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 2026-10-23-5347695500ed. 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.

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 2026-10-23-5347695500ed
Property and statementOriginal source and locationReview and provenance

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

How to read the evidence

Sources and history

Release 2026-10-23-5347695500ed · Record review: source checked

1 source record and release historyDownload this release (gzip)
Technical metadata and extraction receipts

Stable ID: pheno-20261022-qin2025-protocol-downsampling-rmse

areas
cells-tissues
contexts
research
protocol
Training sets are subsampled so that Gene Occurrence = 2N/M (N training combinations, M unique genes) is about 4, 10 or 20; 10% of each dataset is used for validation and 30% for testing. Root mean squared error of predicted viability on the test pairs.
version
Qin et al. 2025, PMLR 267, Sections 4.2 and 5.2; Table 1; Appendix Table 6
metric
root-mean-squared-error
metric direction
lower
unit
unitless
unit detail
printed x 10^-2 (column header 'Dataset RMSE (×10−2)'), RMSE of gene-level log-fold-change viability
replicates
Three cross-fold replicates; standard error printed
task
Predict the viability effect of unseen two-gene perturbations
limitations
Training sizes per dataset for each gene frequency are not printed with Table 1; Table 9 lists an approximate training-set size per average times seen, which may be the same schedule.; This is prediction accuracy over all held-out pairs, not selection of pairs to test.
source locator
Section 4.2; Section 5.2; Table 1 caption; Table 6 caption
publication status
peer_reviewed
review
method: source-hash-verification; scope-review; reviewer: claude; reviewer note: Claude (Opus 5.5) review agent, independent of the extraction agent; no human review claimed; date: 2026-10-22; artifact sha256: 983b2da8aed78f0aced27053301c15f0f67ebd0fac706435bb6800e309dd64eb; retrieval url: https://raw.githubusercontent.com/mlresearch/v267/main/assets/qin25g/qin25g.pdf; method note: Re-hashed the PDF in the scratchpad (SHA-256 matches the source record). Read the cited sections in the pdftotext layer of the pinned PDF.; note: Fields checked against the cited locator; nothing found that the source does not state.
Related records

Suggest a correction