Protocol

Held-out CRISPR enhancer-gene benchmark, weighted metrics, five cell types (Gschwind et al. 2026 Supplementary Table 3)

Enhancer-gene link prediction against CRISPR perturbation outcomes outside the training data.

Overview

Enhancer-gene link prediction against CRISPR perturbation outcomes outside the training data.

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

6 recorded evaluations, 18 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

6 evaluations · 18 results. Different protocols are not a single leaderboard.

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: ABC_A=DNase, C=Average ENCODE Hi-C (Gschwind et al. 2026)	Protocol: Held-out CRISPR enhancer-gene benchmark, weighted metrics, five cell types (Gschwind et al. 2026 Supplementary Table 3) Dataset: Held-out CRISPR enhancer-gene pairs in five cell types (Gschwind et al. 2026)	0.465 auprc fraction · higher Uncertainty: 95% CI 0.378228901913079 to 0.541415553632906 Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Methods, coverage and source ABC_A=DNase, C=Average ENCODE Hi-C on held-out CRISPR pairs, five cell types regulatory-variant-20261009-protocol-gschwind2026-heldout-weighted Aggregation: Not reported An encyclopedia of human enhancer-gene regulatory interactions; Gschwind et al. 2026, Supplementary Table 3 · Supplementary Table 3 'Held-out benchmarks', row 2 (D2:F2); Dataset 'Held-out', Predictor 'ABC_A=DNase, C=Average ENCODE Hi-C', metric 'Weighted AUPRC'
Configuration: ABC_A=DNase, C=Average ENCODE Hi-C (Gschwind et al. 2026)	Protocol: Held-out CRISPR enhancer-gene benchmark, weighted metrics, five cell types (Gschwind et al. 2026 Supplementary Table 3) Dataset: Held-out CRISPR enhancer-gene pairs in five cell types (Gschwind et al. 2026)	0.437 precision fraction · higher Uncertainty: 95% CI 0.375298718902513 to 0.497269228021238 Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Methods, coverage and source ABC_A=DNase, C=Average ENCODE Hi-C on held-out CRISPR pairs, five cell types regulatory-variant-20261009-protocol-gschwind2026-heldout-weighted Aggregation: Not reported An encyclopedia of human enhancer-gene regulatory interactions; Gschwind et al. 2026, Supplementary Table 3 · Supplementary Table 3 'Held-out benchmarks', row 8 (D8:F8); Dataset 'Held-out', Predictor 'ABC_A=DNase, C=Average ENCODE Hi-C', metric 'Weighted precision at threshold'
Configuration: ABC_A=DNase, C=Average ENCODE Hi-C (Gschwind et al. 2026)	Protocol: Held-out CRISPR enhancer-gene benchmark, weighted metrics, five cell types (Gschwind et al. 2026 Supplementary Table 3) Dataset: Held-out CRISPR enhancer-gene pairs in five cell types (Gschwind et al. 2026)	0.619 recall fraction · higher Uncertainty: 95% CI 0.547527036355901 to 0.688813573106359 Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Methods, coverage and source ABC_A=DNase, C=Average ENCODE Hi-C on held-out CRISPR pairs, five cell types regulatory-variant-20261009-protocol-gschwind2026-heldout-weighted Aggregation: Not reported An encyclopedia of human enhancer-gene regulatory interactions; Gschwind et al. 2026, Supplementary Table 3 · Supplementary Table 3 'Held-out benchmarks', row 14 (D14:F14); Dataset 'Held-out', Predictor 'ABC_A=DNase, C=Average ENCODE Hi-C', metric 'Weighted recall at threshold'
Configuration: Distance to TSS (Gschwind et al. 2026)	Protocol: Held-out CRISPR enhancer-gene benchmark, weighted metrics, five cell types (Gschwind et al. 2026 Supplementary Table 3) Dataset: Held-out CRISPR enhancer-gene pairs in five cell types (Gschwind et al. 2026)	0.363 auprc fraction · higher Uncertainty: 95% CI 0.281241216228264 to 0.437896240213437 Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Methods, coverage and source Distance to TSS on held-out CRISPR pairs, five cell types regulatory-variant-20261009-protocol-gschwind2026-heldout-weighted Aggregation: Not reported An encyclopedia of human enhancer-gene regulatory interactions; Gschwind et al. 2026, Supplementary Table 3 · Supplementary Table 3 'Held-out benchmarks', row 6 (D6:F6); Dataset 'Held-out', Predictor 'Distance to TSS', metric 'Weighted AUPRC'
Configuration: Distance to TSS (Gschwind et al. 2026)	Protocol: Held-out CRISPR enhancer-gene benchmark, weighted metrics, five cell types (Gschwind et al. 2026 Supplementary Table 3) Dataset: Held-out CRISPR enhancer-gene pairs in five cell types (Gschwind et al. 2026)	0.245 precision fraction · higher Uncertainty: 95% CI 0.202048095028741 to 0.288268461392336 Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Methods, coverage and source Distance to TSS on held-out CRISPR pairs, five cell types regulatory-variant-20261009-protocol-gschwind2026-heldout-weighted Aggregation: Not reported An encyclopedia of human enhancer-gene regulatory interactions; Gschwind et al. 2026, Supplementary Table 3 · Supplementary Table 3 'Held-out benchmarks', row 12 (D12:F12); Dataset 'Held-out', Predictor 'Distance to TSS', metric 'Weighted precision at threshold'
Configuration: Distance to TSS (Gschwind et al. 2026)	Protocol: Held-out CRISPR enhancer-gene benchmark, weighted metrics, five cell types (Gschwind et al. 2026 Supplementary Table 3) Dataset: Held-out CRISPR enhancer-gene pairs in five cell types (Gschwind et al. 2026)	0.584 recall fraction · higher Uncertainty: 95% CI 0.511553237372335 to 0.655007748643397 Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Methods, coverage and source Distance to TSS on held-out CRISPR pairs, five cell types regulatory-variant-20261009-protocol-gschwind2026-heldout-weighted Aggregation: Not reported An encyclopedia of human enhancer-gene regulatory interactions; Gschwind et al. 2026, Supplementary Table 3 · Supplementary Table 3 'Held-out benchmarks', row 18 (D18:F18); Dataset 'Held-out', Predictor 'Distance to TSS', metric 'Weighted recall at threshold'
Configuration: ENCODE-rE2G (Gschwind et al. 2026)	Protocol: Held-out CRISPR enhancer-gene benchmark, weighted metrics, five cell types (Gschwind et al. 2026 Supplementary Table 3) Dataset: Held-out CRISPR enhancer-gene pairs in five cell types (Gschwind et al. 2026)	0.556 auprc fraction · higher Uncertainty: 95% CI 0.467852411496632 to 0.631223913857611 Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Methods, coverage and source ENCODE-rE2G on held-out CRISPR pairs, five cell types regulatory-variant-20261009-protocol-gschwind2026-heldout-weighted Aggregation: Not reported An encyclopedia of human enhancer-gene regulatory interactions; Gschwind et al. 2026, Supplementary Table 3 · Supplementary Table 3 'Held-out benchmarks', row 3 (D3:F3); Dataset 'Held-out', Predictor 'ENCODE-rE2G', metric 'Weighted AUPRC'
Configuration: ENCODE-rE2G (Gschwind et al. 2026)	Protocol: Held-out CRISPR enhancer-gene benchmark, weighted metrics, five cell types (Gschwind et al. 2026 Supplementary Table 3) Dataset: Held-out CRISPR enhancer-gene pairs in five cell types (Gschwind et al. 2026)	0.544 precision fraction · higher Uncertainty: 95% CI 0.473562214144615 to 0.614971149178317 Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Methods, coverage and source ENCODE-rE2G on held-out CRISPR pairs, five cell types regulatory-variant-20261009-protocol-gschwind2026-heldout-weighted Aggregation: Not reported An encyclopedia of human enhancer-gene regulatory interactions; Gschwind et al. 2026, Supplementary Table 3 · Supplementary Table 3 'Held-out benchmarks', row 9 (D9:F9); Dataset 'Held-out', Predictor 'ENCODE-rE2G', metric 'Weighted precision at threshold'
Configuration: ENCODE-rE2G (Gschwind et al. 2026)	Protocol: Held-out CRISPR enhancer-gene benchmark, weighted metrics, five cell types (Gschwind et al. 2026 Supplementary Table 3) Dataset: Held-out CRISPR enhancer-gene pairs in five cell types (Gschwind et al. 2026)	0.599 recall fraction · higher Uncertainty: 95% CI 0.526506289907121 to 0.670411257697732 Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Methods, coverage and source ENCODE-rE2G on held-out CRISPR pairs, five cell types regulatory-variant-20261009-protocol-gschwind2026-heldout-weighted Aggregation: Not reported An encyclopedia of human enhancer-gene regulatory interactions; Gschwind et al. 2026, Supplementary Table 3 · Supplementary Table 3 'Held-out benchmarks', row 15 (D15:F15); Dataset 'Held-out', Predictor 'ENCODE-rE2G', metric 'Weighted recall at threshold'
Configuration: Correlation_E-P DNase-seq signal (Gschwind et al. 2026)	Protocol: Held-out CRISPR enhancer-gene benchmark, weighted metrics, five cell types (Gschwind et al. 2026 Supplementary Table 3) Dataset: Held-out CRISPR enhancer-gene pairs in five cell types (Gschwind et al. 2026)	0.211 auprc fraction · higher Uncertainty: 95% CI 0.155299843477219 to 0.267454116811189 Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Methods, coverage and source Correlation_E-P DNase-seq signal on held-out CRISPR pairs, five cell types regulatory-variant-20261009-protocol-gschwind2026-heldout-weighted Aggregation: Not reported An encyclopedia of human enhancer-gene regulatory interactions; Gschwind et al. 2026, Supplementary Table 3 · Supplementary Table 3 'Held-out benchmarks', row 7 (D7:F7); Dataset 'Held-out', Predictor 'Correlation_E-P DNase-seq signal', metric 'Weighted AUPRC'
Configuration: Correlation_E-P DNase-seq signal (Gschwind et al. 2026)	Protocol: Held-out CRISPR enhancer-gene benchmark, weighted metrics, five cell types (Gschwind et al. 2026 Supplementary Table 3) Dataset: Held-out CRISPR enhancer-gene pairs in five cell types (Gschwind et al. 2026)	0.112 precision fraction · higher Uncertainty: 95% CI 0.09490444 to 0.130640831462308 Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Methods, coverage and source Correlation_E-P DNase-seq signal on held-out CRISPR pairs, five cell types regulatory-variant-20261009-protocol-gschwind2026-heldout-weighted Aggregation: Not reported An encyclopedia of human enhancer-gene regulatory interactions; Gschwind et al. 2026, Supplementary Table 3 · Supplementary Table 3 'Held-out benchmarks', row 13 (D13:F13); Dataset 'Held-out', Predictor 'Correlation_E-P DNase-seq signal', metric 'Weighted precision at threshold'
Configuration: Correlation_E-P DNase-seq signal (Gschwind et al. 2026)	Protocol: Held-out CRISPR enhancer-gene benchmark, weighted metrics, five cell types (Gschwind et al. 2026 Supplementary Table 3) Dataset: Held-out CRISPR enhancer-gene pairs in five cell types (Gschwind et al. 2026)	0.737 recall fraction · higher Uncertainty: 95% CI 0.669820594780111 to 0.79971637 Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Methods, coverage and source Correlation_E-P DNase-seq signal on held-out CRISPR pairs, five cell types regulatory-variant-20261009-protocol-gschwind2026-heldout-weighted Aggregation: Not reported An encyclopedia of human enhancer-gene regulatory interactions; Gschwind et al. 2026, Supplementary Table 3 · Supplementary Table 3 'Held-out benchmarks', row 19 (D19:F19); Dataset 'Held-out', Predictor 'Correlation_E-P DNase-seq signal', metric 'Weighted recall at threshold'
Configuration: EpiMap (Gschwind et al. 2026)	Protocol: Held-out CRISPR enhancer-gene benchmark, weighted metrics, five cell types (Gschwind et al. 2026 Supplementary Table 3) Dataset: Held-out CRISPR enhancer-gene pairs in five cell types (Gschwind et al. 2026)	0.268 auprc fraction · higher Uncertainty: 95% CI 0.202767496369732 to 0.342972872868445 Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Methods, coverage and source EpiMap on held-out CRISPR pairs, five cell types regulatory-variant-20261009-protocol-gschwind2026-heldout-weighted Aggregation: Not reported An encyclopedia of human enhancer-gene regulatory interactions; Gschwind et al. 2026, Supplementary Table 3 · Supplementary Table 3 'Held-out benchmarks', row 5 (D5:F5); Dataset 'Held-out', Predictor 'EpiMap', metric 'Weighted AUPRC'
Configuration: EpiMap (Gschwind et al. 2026)	Protocol: Held-out CRISPR enhancer-gene benchmark, weighted metrics, five cell types (Gschwind et al. 2026 Supplementary Table 3) Dataset: Held-out CRISPR enhancer-gene pairs in five cell types (Gschwind et al. 2026)	0.292 precision fraction · higher Uncertainty: 95% CI 0.244244656893669 to 0.342593944957384 Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Methods, coverage and source EpiMap on held-out CRISPR pairs, five cell types regulatory-variant-20261009-protocol-gschwind2026-heldout-weighted Aggregation: Not reported An encyclopedia of human enhancer-gene regulatory interactions; Gschwind et al. 2026, Supplementary Table 3 · Supplementary Table 3 'Held-out benchmarks', row 11 (D11:F11); Dataset 'Held-out', Predictor 'EpiMap', metric 'Weighted precision at threshold'
Configuration: EpiMap (Gschwind et al. 2026)	Protocol: Held-out CRISPR enhancer-gene benchmark, weighted metrics, five cell types (Gschwind et al. 2026 Supplementary Table 3) Dataset: Held-out CRISPR enhancer-gene pairs in five cell types (Gschwind et al. 2026)	0.545 recall fraction · higher Uncertainty: 95% CI 0.471122984227133 to 0.616597040982763 Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Methods, coverage and source EpiMap on held-out CRISPR pairs, five cell types regulatory-variant-20261009-protocol-gschwind2026-heldout-weighted Aggregation: Not reported An encyclopedia of human enhancer-gene regulatory interactions; Gschwind et al. 2026, Supplementary Table 3 · Supplementary Table 3 'Held-out benchmarks', row 17 (D17:F17); Dataset 'Held-out', Predictor 'EpiMap', metric 'Weighted recall at threshold'
Configuration: EPIraction (Gschwind et al. 2026)	Protocol: Held-out CRISPR enhancer-gene benchmark, weighted metrics, five cell types (Gschwind et al. 2026 Supplementary Table 3) Dataset: Held-out CRISPR enhancer-gene pairs in five cell types (Gschwind et al. 2026)	0.383 auprc fraction · higher Uncertainty: 95% CI 0.302549537433103 to 0.455764606497437 Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Methods, coverage and source EPIraction on held-out CRISPR pairs, five cell types regulatory-variant-20261009-protocol-gschwind2026-heldout-weighted Aggregation: Not reported An encyclopedia of human enhancer-gene regulatory interactions; Gschwind et al. 2026, Supplementary Table 3 · Supplementary Table 3 'Held-out benchmarks', row 4 (D4:F4); Dataset 'Held-out', Predictor 'EPIraction', metric 'Weighted AUPRC'
Configuration: EPIraction (Gschwind et al. 2026)	Protocol: Held-out CRISPR enhancer-gene benchmark, weighted metrics, five cell types (Gschwind et al. 2026 Supplementary Table 3) Dataset: Held-out CRISPR enhancer-gene pairs in five cell types (Gschwind et al. 2026)	0.347 precision fraction · higher Uncertainty: 95% CI 0.292942536891562 to 0.401809875530837 Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Methods, coverage and source EPIraction on held-out CRISPR pairs, five cell types regulatory-variant-20261009-protocol-gschwind2026-heldout-weighted Aggregation: Not reported An encyclopedia of human enhancer-gene regulatory interactions; Gschwind et al. 2026, Supplementary Table 3 · Supplementary Table 3 'Held-out benchmarks', row 10 (D10:F10); Dataset 'Held-out', Predictor 'EPIraction', metric 'Weighted precision at threshold'
Configuration: EPIraction (Gschwind et al. 2026)	Protocol: Held-out CRISPR enhancer-gene benchmark, weighted metrics, five cell types (Gschwind et al. 2026 Supplementary Table 3) Dataset: Held-out CRISPR enhancer-gene pairs in five cell types (Gschwind et al. 2026)	0.596 recall fraction · higher Uncertainty: 95% CI 0.525947724997951 to 0.666054834012973 Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Methods, coverage and source EPIraction on held-out CRISPR pairs, five cell types regulatory-variant-20261009-protocol-gschwind2026-heldout-weighted Aggregation: Not reported An encyclopedia of human enhancer-gene regulatory interactions; Gschwind et al. 2026, Supplementary Table 3 · Supplementary Table 3 'Held-out benchmarks', row 16 (D16:F16); Dataset 'Held-out', Predictor 'EPIraction', metric 'Weighted recall at threshold'

Source checking is not independent reproduction. Release 2026-10-10-7fcc3e48a123.

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: 6

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 2026-10-10-7fcc3e48a123. 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 2026-10-10-7fcc3e48a123
Property and statement	Original source and location	Review and provenance

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

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

Sources and history

Release 2026-10-10-7fcc3e48a123 · Record review: source checked

2 source records and release history

An encyclopedia of human enhancer-gene regulatory interactions · Original source · Nature 657(8130):179, published online 2026-07-15; PMC13471189 full-text XML
Gschwind et al. 2026, Supplementary Table 3 · Original source · Supplementary_Table_3.xlsx in 41586_2026_10781_MOESM3_ESM.zip (folder 2023-11-20318B-s3) inside the Europe PMC supplementaryFiles zip

Download this release (gzip)

Technical metadata and extraction receipts

Stable ID: regulatory-variant-20261009-protocol-gschwind2026-heldout-weighted

areas: dna-genomes
contexts: research
protocol: Overlap CRISPR-tested elements with each predictor's elements (matched by gene symbol; sum or max aggregation per Supplementary Table 1; unmatched pairs get the minimum score). Weighted precision-recall (yardstick 1.2.0), each pair weighted by its estimated probability of a direct cis-regulatory effect; weighted AUPRC, and weighted precision and recall at the predictor threshold, with 95% bootstrap intervals (10,000 iterations).
version: Supplementary Table 3 sheet 'Held-out benchmarks', Dataset 'Held-out'
limitations: Developer comparison: ENCODE-rE2G and ABC are the authors' models, and EPIraction and EpiMap come from co-authors' groups; all six rows are author_reported, and the baselines were computed by the authors.; Whole-element CRISPR perturbation, not allele editing; a regulatory link does not establish a causal disease role.; Each pair is weighted by the authors' estimate of its probability of being a direct effect (Methods 'Calculating direct and indirect effect rates').; Precision and recall 'at threshold' use each predictor's threshold set at 70% recall on the combined K562 CRISPR data (Methods; Supplementary Table 3 legend).; Held out: pairs present in the K562 training data were removed (Methods 'Creating the combined held-out CRISPR dataset'), but K562 pairs from other screens remain, and model design and thresholds were chosen on K562.; 4,378 pairs and 190 positives (157.39 weighted) pooled over five cell types; no per-cell-type values in this sheet.
source locator: Supplementary Table 3 'Held-out benchmarks' rows 2-19; Methods 'CRISPR benchmark'

Related records

uses data: Held-out CRISPR enhancer-gene pairs in five cell types (Gschwind et al. 2026)
assessment: ABC_A=DNase, C=Average ENCODE Hi-C on held-out CRISPR pairs, five cell types
assessment: Distance to TSS on held-out CRISPR pairs, five cell types
assessment: ENCODE-rE2G on held-out CRISPR pairs, five cell types
assessment: Correlation_E-P DNase-seq signal on held-out CRISPR pairs, five cell types
assessment: EpiMap on held-out CRISPR pairs, five cell types
assessment: EPIraction on held-out CRISPR pairs, five cell types
assessed by: Select regulatory variants and genes for functional follow-up

Suggest a correction