Benchmark

Static GRN reconstruction from single-cell foundation models (Ye et al. 2026)

Evidence concern: excluded from comparisons

Table 1 labels its middle reference column 'ChIP-seq' and the caption names cell-type-specific ChIP-seq, but its values equal the 'Nonspecific ChIP-seq' mean and std rows of Tables 4 and 6 (for example scGPT costok EPR 4.211±1.703), and Appendix B says cell-type-specific ChIP-seq is excluded from the main comparison. Table 1 is therefore not extracted; Tables 4-7 carry the reference labels.Benchmarking Static Gene Regulatory Network Reconstruction and Dynamic Transition Probing in Single-Cell Foundation Models · Table 1 caption and column header; Tables 4 and 6, 'Nonspecific ChIP-seq' mean rows; Appendix B text before Table 4

Table 6 is titled 'AUPR' and Table 7 'AUPRC', but both print ratios over a random predictor (values above 1), matching the 'AUPRC ratio' of Table 1; both are recorded as auprc-ratio.Benchmarking Static Gene Regulatory Network Reconstruction and Dynamic Transition Probing in Single-Cell Foundation Models · Table 6 and Table 7 captions; Table 1 caption

Table 7 (TFs+1000 AUPRC) repeats Table 6 (TFs+500 AUPR) exactly in the GENIE3 column of the STRING, Cell-type-specific ChIP-seq and Nonspecific ChIP-seq blocks: all 18 dataset values and the three mean and std rows are identical (for example STRING hESC 1.6241, mean 3.301, std 2.244). Only the OmniPath block differs. Every other column, and GENIE3 in the EPR Tables 4 and 5, differs between the two gene sets, and the Table 6 values match Table 1. These 21 Table 7 GENIE3 values are likely a copy of the TFs+500 values and should not be compared.Benchmarking Static Gene Regulatory Network Reconstruction and Dynamic Transition Probing in Single-Cell Foundation Models · Table 7, column group 'ML', column 'GENIE3', blocks 'STRING', 'Cell-type-specific' and 'Nonspecific'; Table 6, same cells; Table 1, GENIE3 AUPRC ratio

Zero-shot GRN reconstruction from six single-cell foundation models with three extraction strategies (token-embedding cosine, hidden-state cosine, attention) against GENIE3, DeepSEM and GRNBoost, on six BEELINE scRNA-seq datasets and four reference network types.

Results are recorded on the linked protocols, tasks or configurations listed below. Their settings and evaluations are kept separate.

8 protocols with 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.

Results on linked records

This benchmark has no results of its own in this release. 8 linked records have results. They are not assigned to the underlying benchmark: each record keeps its own settings and evaluations, and its scores are on its own page.

Ye et al. 2026 GRN reconstruction, TFs+1000, cell-type-specific ChIP-seq reference Protocol · part of this benchmark · 147 evaluations · 294 results
Ye et al. 2026 GRN reconstruction, TFs+1000, non-specific ChIP-seq reference Protocol · part of this benchmark · 147 evaluations · 294 results
Ye et al. 2026 GRN reconstruction, TFs+1000, OmniPath reference Protocol · part of this benchmark · 147 evaluations · 294 results
Ye et al. 2026 GRN reconstruction, TFs+1000, STRING reference Protocol · part of this benchmark · 147 evaluations · 294 results
Ye et al. 2026 GRN reconstruction, TFs+500, cell-type-specific ChIP-seq reference Protocol · part of this benchmark · 147 evaluations · 294 results
Ye et al. 2026 GRN reconstruction, TFs+500, non-specific ChIP-seq reference Protocol · part of this benchmark · 147 evaluations · 294 results
Ye et al. 2026 GRN reconstruction, TFs+500, OmniPath reference Protocol · part of this benchmark · 147 evaluations · 294 results
Ye et al. 2026 GRN reconstruction, TFs+500, STRING reference Protocol · part of this benchmark · 147 evaluations · 294 results

Methods and evaluation design

Procedure, tasks and evaluated configurations

Additional methodological details, tasks and evaluated configurations have not yet been extracted into a structured profile for this benchmark. Consult the linked sources and evaluation records below.

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

Baseline status by linked protocol

Ye et al. 2026 GRN reconstruction, TFs+1000, cell-type-specific ChIP-seq reference · 0/2 roles measured
Ye et al. 2026 GRN reconstruction, TFs+1000, non-specific ChIP-seq reference · 0/2 roles measured
Ye et al. 2026 GRN reconstruction, TFs+1000, OmniPath reference · 0/2 roles measured
Ye et al. 2026 GRN reconstruction, TFs+1000, STRING reference · 0/2 roles measured
Ye et al. 2026 GRN reconstruction, TFs+500, cell-type-specific ChIP-seq reference · 0/2 roles measured
Ye et al. 2026 GRN reconstruction, TFs+500, non-specific ChIP-seq reference · 0/2 roles measured
Ye et al. 2026 GRN reconstruction, TFs+500, OmniPath reference · 0/2 roles measured
Ye et al. 2026 GRN reconstruction, TFs+500, STRING reference · 0/2 roles measured

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

1 source record and release history

Benchmarking Static Gene Regulatory Network Reconstruction and Dynamic Transition Probing in Single-Cell Foundation Models · Original source · bioRxiv preprint version 1, posted 2026-05-20

Download this release (gzip)

Technical metadata and extraction receipts

Stable ID: lit-cells-networks-20261011-ye2026grn-benchmark-static-grn-fm

areas: biological-networks; single-cell
contexts: research
entity level: suite
version: bioRxiv v1 (2026-05-20)
task: Gene regulatory network inference from single-cell expression
scope note: Uses BEELINE datasets and preprocessing with TF-prioritised gene sets (TFs+500, TFs+1000); edges restricted to TF-target candidates and ranked by absolute weight.
source locator: Sections 2.1-2.2; Appendix A-B; Tables 4-7

Related records

Suggest a correction