Result

Hits at 100 (OGBl-ppa PPI link prediction) of GraphSAGE with ESM-2 protein embeddings (sequence) (PPI embeddings study) on OGBl-ppa PPI graph (embedding study)

0.242 Hits at 100 (OGBl-ppa PPI link prediction)

Evidence concern: excluded from comparisons

The text's summary does not match the tables: it lists ProtT5 among structure embeddings while the tables print ProstT5; it reports a GraphSAGE with DistMult result (+38% Hits@100 on HuRI) although DistMult is not a column; and the stated gains (for example ESM-2 with GraphSAGE +6% Hits@100 on OGBl-ppa) cannot be reproduced from the printed table values (GS random 0.0104, ESM-2 0.2421). Values are kept as printed; comparisons should rely on the tables only after a reviewer has judged this. Random-feature baselines on OGBl-ppa (GCN 0.0588, GS 0.0104 Hits@100) look far below the public OGB leaderboard values the authors say they reproduced (Supplementary Table S2, not retrieved; reviewer did not verify the leaderboard values).The role of protein embeddings for protein-protein interaction prediction with graph neural networks · Results and discussion, first paragraph; Tables 5 and 6

Methods

Tested configuration: GraphSAGE with ESM-2 protein embeddings (sequence) (PPI embeddings study)
Protocol: OGBl-ppa PPI link prediction, Hits@100 (Table 5)
Dataset: OGBl-ppa PPI graph (embedding study)
Procedure: OGBl-ppa PPI link prediction, Hits@100 (Table 5)
Evaluation: GraphSAGE / ESM-2 on OGBl-ppa (Hits@100)
Coverage: Not reported scored / Not reported eligible
Uncertainty: Not reported by the source: Only a table-wide maximum SD is printed (Table 5/6 footnote)
Evidence: Independent external evaluation · source checkedThe role of protein embeddings for protein-protein interaction prediction with graph neural networks · Table 5, block 'OGBl-ppa', row 'GS', column 'ESM-2'

A source-checked result verifies the numerical transcription, not every model or protocol detail. Evaluation metadata: source checked. Source checked does not mean independently reproduced.

Reproduction

Split: Benchmark test edges
Adaptation: Trained by the study authors
Scoring implementation: Hits@100 as printed

No execution recipe has been verified for this exact configuration and evaluation. A benchmark's general instructions may use different inputs, splits or model settings.

Reproducing this published result requires matching its model configuration, data, split and scorer. Source checking or a successful smoke test does not establish score reproduction.

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 2026-10-22-1b7cf027e4f6
Property and statement	Original source and location	Review and provenance
Reported result 0.2421 Individual claims	The role of protein embeddings for protein-protein interaction prediction with graph neural networks Original source ↗ Table 5, block 'OGBl-ppa', row 'GS', column 'ESM-2' Version: Briefings in Bioinformatics 2026, bbag313; PMC13280947 full-text XML Retrieved: 2026-10-11T21:27:23Z	source checked ["source-hash-verification","deterministic-table-parse","independent-cell-check","rendered-source-visual-check"] · 2026-10-11 independent paper Source has a recorded evidence concern. Consult its source page before using the claim. Audit details Extraction: Parsed the <td> cells of the pinned PMC XML with extract/extract_ppiemb.py; header rows and row lengths asserted. Table 5 body row 4. Independent review 2026-10-11: Re-downloaded the PMC full-text XML from the recorded Europe PMC URL and matched its SHA-256. Parsed all 308 cells (2 tables x 2 datasets x 7 rows x 11 columns) with an independent parser, mapping every result through its evaluation, configuration and protocol: all match and no cell was missed. The misaligned HuRI N-GCN row is laid out the same way in the XML and the PMC HTML, so reading its eleven values in column order is the only consistent reading. Read 42 values covering both tables, both datasets, all seven rows and all eleven columns from the PMC HTML rendering: 42 of 42 match. The three prose-versus-table discrepancies were confirmed, and none of the six percentage gains in the first Results paragraph can be derived from Table 5. Field: `attributes.printed_value` Source artifact SHA-256: `360d4f17b779e07f170c3f0c3755bfbfddd064831bf5a81d2642a4cfffd4866c` Hash scope: Hash scope not separately documented; inspect source record Inspected artifact Extraction artifact SHA-256: `360d4f17b779e07f170c3f0c3755bfbfddd064831bf5a81d2642a4cfffd4866c` Extraction artifact

How to read the evidence

Sources and history

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

1 source record and release history

The role of protein embeddings for protein-protein interaction prediction with graph neural networks · Original source · Briefings in Bioinformatics 2026, bbag313; PMC13280947 full-text XML

Download this release (gzip)

Technical metadata and extraction receipts

Stable ID: lit-interactions-20261011-ppiemb-result-gs-esm-2-ogbl-ppa-hits-100

metric: hits-at-100
metric direction: higher
unit: fraction
printed value: 0.2421
source locator: Table 5, block 'OGBl-ppa', row 'GS', column 'ESM-2'
review: method: source-hash-verification; deterministic-table-parse; independent-cell-check; rendered-source-visual-check; method note: Re-downloaded the PMC full-text XML from the recorded Europe PMC URL and matched its SHA-256. Parsed all 308 cells (2 tables x 2 datasets x 7 rows x 11 columns) with an independent parser, mapping every result through its evaluation, configuration and protocol: all match and no cell was missed. The misaligned HuRI N-GCN row is laid out the same way in the XML and the PMC HTML, so reading its eleven values in column order is the only consistent reading. Read 42 values covering both tables, both datasets, all seven rows and all eleven columns from the PMC HTML rendering: 42 of 42 match. The three prose-versus-table discrepancies were confirmed, and none of the six percentage gains in the first Results paragraph can be derived from Table 5.; reviewer: claude; reviewer note: Claude review agent reviewer-mi-2, independent of the extractor lit-microbial-interactions; no human review claimed; date: 2026-10-11; artifact sha256: 360d4f17b779e07f170c3f0c3755bfbfddd064831bf5a81d2642a4cfffd4866c; retrieval url: https://www.ebi.ac.uk/europepmc/webservices/rest/PMC13280947/fullTextXML; note: Extraction: Parsed the <td> cells of the pinned PMC XML with extract/extract_ppiemb.py; header rows and row lengths asserted. Table 5 body row 4. Independent review 2026-10-11: Re-downloaded the PMC full-text XML from the recorded Europe PMC URL and matched its SHA-256. Parsed all 308 cells (2 tables x 2 datasets x 7 rows x 11 columns) with an independent parser, mapping every result through its evaluation, configuration and protocol: all match and no cell was missed. The misaligned HuRI N-GCN row is laid out the same way in the XML and the PMC HTML, so reading its eleven values in column order is the only consistent reading. Read 42 values covering both tables, both datasets, all seven rows and all eleven columns from the PMC HTML rendering: 42 of 42 match. The three prose-versus-table discrepancies were confirmed, and none of the six percentage gains in the first Results paragraph can be derived from Table 5.
numeric value: 0.2421
metric qualifier: OGBl-ppa PPI link prediction
missing metadata: uncertainty: reason: unreported; note: Only a table-wide maximum SD is printed (Table 5/6 footnote)

Related records

evaluation: GraphSAGE / ESM-2 on OGBl-ppa (Hits@100)

Suggest a correction