Comparative evaluation of the prediction accuracy of AlphaFold and ESMFold for monomeric and dimeric proteins
NAR Genomics and Bioinformatics · 2026 · Version of record, NAR Genomics and Bioinformatics 2026 (PMC12809598 JATS full text) · Peer-reviewed
What this source contributed
Records in this release that cite this source. A record can cite several sources.
Results
Values as the source printed them. Compare results only under compatible protocols, datasets and metrics.
Claims
Statements this source supports, each with where in the source it was checked.
- printed_value_conflict: lit-structure-20261011-afesmfold-source-mahtha2026: Printed value conflict The Abstract prints one dimer figure for both tools: 'AlphaFold2 and AlphaFold3 correctly predicted 88% of monomeric structures and 77% of dimeric proteins.' Table 2 gives AlphaFold2 361 high, 297 medium and 105 acceptable quality dimers out of 994, and AlphaFold3 346, 306 and 93. The Results print these totals as 763 (76.8%) for AlphaFold2 and 745 (75%) for AlphaFold3, and the Discussion as 76.8% and 75%. The Abstract's 77% matches AlphaFold2 only; for AlphaFold3 the Table 2 counts give 745/994, 74.9% to one decimal (computed here, not printed). (Abstract sentence 3; Table 2, AlphaFold2 and AlphaFold3 rows 'High quality', 'Medium quality', 'Acceptable quality', column 'No.'; Results section 'Performance of AlphaFold for dimeric structure prediction' paragraphs 1 and 3; Discussion paragraph 3)
- printed_value_conflict: lit-structure-20261011-afesmfold-source-mahtha2026: Printed value conflict The Abstract prints ESMFold monomer accuracy as 76%: 'On the other hand, ESMFold accurately predicted 76% of the monomeric proteins and 41% of the dimeric proteins.' The Discussion prints 77%: 'For monomeric proteins, AlphaFold2 and AlphaFold3 exhibited comparable performance, each accurately predicting 87%–88% of the structures, while ESMFold achieved 77% accuracy.' Table 1 and the Results give ESMFold 832 accurate predictions (50%) and 430 with domain movement (26%) out of 1,666 monomers. The paper does not print the sum; 832 + 430 = 1,262 of 1,666 is 75.75% (computed here, not printed), which rounds to the Abstract's 76%, not to the Discussion's 77%. (Abstract sentence 4; Discussion paragraph 1, sentence 6; Results section 'Performance of ESMFold for monomer structure prediction' paragraph 1, sentence 1; Table 1, ESMFold rows 'Accurate prediction' and 'Domain movement', column 'No.')
- Relevance of dimeric predictive-class counts to "Choose structural hypotheses to guide experiments": Assessed by Predictive-class counts for dimeric predictions, all structures in the class (Table 2, column 'No.') (Table 2, column 'No.')
- Relevance of monomeric predictive-class counts to "Choose structural hypotheses to guide experiments": Assessed by Predictive-class counts for monomeric predictions, all structures in the class (Table 1, column 'No.') (Table 1, column 'No.')
Models, benchmarks and other records
Models (1)
Configurations (3)
Benchmarks (1)
Protocols (18)
- Predictive-class counts for dimeric predictions, all structures in the class (Table 2, column 'No.')
- Predictive-class counts for dimeric predictions, heterodimeric structures in the class (Table 2, column 'Heterodimer')
- Predictive-class counts for dimeric predictions, homodimeric structures in the class (Table 2, column 'Homodimer')
- Predictive-class counts for dimeric predictions, structures in the class whose experimental reference is electron microscopy (Table 2, column 'EM')
- Predictive-class counts for dimeric predictions, structures in the class whose experimental reference is NMR (Table 2, column 'NMR')
- Predictive-class counts for dimeric predictions, structures in the class whose experimental reference is X-ray diffraction (Table 2, column 'XRD')
- Predictive-class counts for dimeric predictions, structures in the class whose pLDDT is at least 70 (Table 2, column 'pLDDT ≥ 70')
- Predictive-class counts for dimeric predictions, structures in the class with template quality 0 (no homologous pre-2022 structure) (Table 2, column 'TQ 0')
- Predictive-class counts for dimeric predictions, structures in the class with template quality above 40 (Table 2, column 'TQ > 40')
- Predictive-class counts for dimeric predictions, structures in the class with template quality between 1 and 40 (Table 2, column 'TQ 1–40')
- Predictive-class counts for monomeric predictions, all structures in the class (Table 1, column 'No.')
- Predictive-class counts for monomeric predictions, structures in the class whose experimental reference is electron microscopy (Table 1, column 'EM')
- Predictive-class counts for monomeric predictions, structures in the class whose experimental reference is NMR (Table 1, column 'NMR')
- Predictive-class counts for monomeric predictions, structures in the class whose experimental reference is X-ray diffraction (Table 1, column 'XRD')
- Predictive-class counts for monomeric predictions, structures in the class whose pLDDT is at least 70 (Table 1, column 'pLDDT ≥ 70')
- Predictive-class counts for monomeric predictions, structures in the class with template quality 0 (no homologous pre-2022 structure) (Table 1, column 'TQ 0')
- Predictive-class counts for monomeric predictions, structures in the class with template quality above 40 (Table 1, column 'TQ > 40')
- Predictive-class counts for monomeric predictions, structures in the class with template quality between 1 and 40 (Table 1, column 'TQ 1–40')
Datasets (2)
Evaluations (54)
- AlphaFold2 dimeric predictive-class counts, all structures in the class (Table 2)
- AlphaFold2 dimeric predictive-class counts, heterodimeric structures in the class (Table 2)
- AlphaFold2 dimeric predictive-class counts, homodimeric structures in the class (Table 2)
- AlphaFold2 dimeric predictive-class counts, structures in the class whose experimental reference is electron microscopy (Table 2)
- AlphaFold2 dimeric predictive-class counts, structures in the class whose experimental reference is NMR (Table 2)
- AlphaFold2 dimeric predictive-class counts, structures in the class whose experimental reference is X-ray diffraction (Table 2)
- AlphaFold2 dimeric predictive-class counts, structures in the class whose pLDDT is at least 70 (Table 2)
- AlphaFold2 dimeric predictive-class counts, structures in the class with template quality 0 (no homologous pre-2022 structure) (Table 2)
- AlphaFold2 dimeric predictive-class counts, structures in the class with template quality above 40 (Table 2)
- AlphaFold2 dimeric predictive-class counts, structures in the class with template quality between 1 and 40 (Table 2)
- AlphaFold2 monomeric predictive-class counts, all structures in the class (Table 1)
- AlphaFold2 monomeric predictive-class counts, structures in the class whose experimental reference is electron microscopy (Table 1)
- AlphaFold2 monomeric predictive-class counts, structures in the class whose experimental reference is NMR (Table 1)
- AlphaFold2 monomeric predictive-class counts, structures in the class whose experimental reference is X-ray diffraction (Table 1)
- AlphaFold2 monomeric predictive-class counts, structures in the class whose pLDDT is at least 70 (Table 1)
- AlphaFold2 monomeric predictive-class counts, structures in the class with template quality 0 (no homologous pre-2022 structure) (Table 1)
- AlphaFold2 monomeric predictive-class counts, structures in the class with template quality above 40 (Table 1)
- AlphaFold2 monomeric predictive-class counts, structures in the class with template quality between 1 and 40 (Table 1)
- AlphaFold3 dimeric predictive-class counts, all structures in the class (Table 2)
- AlphaFold3 dimeric predictive-class counts, heterodimeric structures in the class (Table 2)
- AlphaFold3 dimeric predictive-class counts, homodimeric structures in the class (Table 2)
- AlphaFold3 dimeric predictive-class counts, structures in the class whose experimental reference is electron microscopy (Table 2)
- AlphaFold3 dimeric predictive-class counts, structures in the class whose experimental reference is NMR (Table 2)
- AlphaFold3 dimeric predictive-class counts, structures in the class whose experimental reference is X-ray diffraction (Table 2)
- AlphaFold3 dimeric predictive-class counts, structures in the class whose pLDDT is at least 70 (Table 2)
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.
26 evidence rows matching the loaded filters
| Property and statement | Original source and location | Review and provenance |
|---|---|---|
| attributes.artifact_sha256 aae321d05839186a7595987d05af6e95aec5ac32b93b35c5cf1126902fe798b8 Source metadata | Comparative evaluation of the prediction accuracy of AlphaFold and ESMFold for monomeric and dimeric proteins No field-specific location recorded Version: Version of record, NAR Genomics and Bioinformatics 2026 (PMC12809598 JATS full text) | catalogued No individual claim review recorded Source has a recorded evidence concern. Consult its source page before using the claim. Audit detailsField: Source artifact SHA-256: Hash scope: Hash scope not separately documented; inspect source record |
| Artifact url https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12809598/fullTextXML Source metadata | Comparative evaluation of the prediction accuracy of AlphaFold and ESMFold for monomeric and dimeric proteins No field-specific location recorded Version: Version of record, NAR Genomics and Bioinformatics 2026 (PMC12809598 JATS full text) | catalogued No individual claim review recorded Source has a recorded evidence concern. Consult its source page before using the claim. Audit detailsField: Source artifact SHA-256: Hash scope: Hash scope not separately documented; inspect source record |
| Doi 10.1093/nargab/lqag002 Source metadata | Comparative evaluation of the prediction accuracy of AlphaFold and ESMFold for monomeric and dimeric proteins No field-specific location recorded Version: Version of record, NAR Genomics and Bioinformatics 2026 (PMC12809598 JATS full text) | catalogued No individual claim review recorded Source has a recorded evidence concern. Consult its source page before using the claim. Audit detailsField: Source artifact SHA-256: Hash scope: Hash scope not separately documented; inspect source record |
Evidence concerns5 values
| Comparative evaluation of the prediction accuracy of AlphaFold and ESMFold for monomeric and dimeric proteins No field-specific location recorded Version: Version of record, NAR Genomics and Bioinformatics 2026 (PMC12809598 JATS full text) | catalogued No individual claim review recorded Source has a recorded evidence concern. Consult its source page before using the claim. Audit detailsField: Source artifact SHA-256: Hash scope: Hash scope not separately documented; inspect source record |
| Licence CC BY-NC 4.0 Source metadata | Comparative evaluation of the prediction accuracy of AlphaFold and ESMFold for monomeric and dimeric proteins No field-specific location recorded Version: Version of record, NAR Genomics and Bioinformatics 2026 (PMC12809598 JATS full text) | catalogued No individual claim review recorded Source has a recorded evidence concern. Consult its source page before using the claim. Audit detailsField: Source artifact SHA-256: Hash scope: Hash scope not separately documented; inspect source record |
Limitations3 values
| Comparative evaluation of the prediction accuracy of AlphaFold and ESMFold for monomeric and dimeric proteins No field-specific location recorded Version: Version of record, NAR Genomics and Bioinformatics 2026 (PMC12809598 JATS full text) | catalogued No individual claim review recorded Source has a recorded evidence concern. Consult its source page before using the claim. Audit detailsField: Source artifact SHA-256: Hash scope: Hash scope not separately documented; inspect source record |
| Media type application/xml Source metadata | Comparative evaluation of the prediction accuracy of AlphaFold and ESMFold for monomeric and dimeric proteins No field-specific location recorded Version: Version of record, NAR Genomics and Bioinformatics 2026 (PMC12809598 JATS full text) | catalogued No individual claim review recorded Source has a recorded evidence concern. Consult its source page before using the claim. Audit detailsField: Source artifact SHA-256: Hash scope: Hash scope not separately documented; inspect source record |
| Publication status peer_reviewed Source metadata | Comparative evaluation of the prediction accuracy of AlphaFold and ESMFold for monomeric and dimeric proteins No field-specific location recorded Version: Version of record, NAR Genomics and Bioinformatics 2026 (PMC12809598 JATS full text) | catalogued No individual claim review recorded Source has a recorded evidence concern. Consult its source page before using the claim. Audit detailsField: Source artifact SHA-256: Hash scope: Hash scope not separately documented; inspect source record |
| Retrieved at 2026-10-11T21:30:00Z Source metadata | Comparative evaluation of the prediction accuracy of AlphaFold and ESMFold for monomeric and dimeric proteins No field-specific location recorded Version: Version of record, NAR Genomics and Bioinformatics 2026 (PMC12809598 JATS full text) | catalogued No individual claim review recorded Source has a recorded evidence concern. Consult its source page before using the claim. Audit detailsField: Source artifact SHA-256: Hash scope: Hash scope not separately documented; inspect source record |
| Review: date 2026-10-11 Source metadata | Comparative evaluation of the prediction accuracy of AlphaFold and ESMFold for monomeric and dimeric proteins No field-specific location recorded Version: Version of record, NAR Genomics and Bioinformatics 2026 (PMC12809598 JATS full text) | catalogued No individual claim review recorded Source has a recorded evidence concern. Consult its source page before using the claim. Audit detailsField: Source artifact SHA-256: Hash scope: Hash scope not separately documented; inspect source record |
Sources and history
Release 2026-10-22-1b7cf027e4f6 · Record review: source checked
Primary source retrieved and hashed for the 2026-10-11 protein and structure literature pass.
Technical metadata and extraction receipts
Stable ID: lit-structure-20261011-afesmfold-source-mahtha2026
- areas
- protein-structure; proteins-complexes
- contexts
- research
- url
- https://pmc.ncbi.nlm.nih.gov/articles/PMC12809598/
- artifact url
- https://www.ebi.ac.uk/europepmc/webservices/rest/PMC12809598/fullTextXML
- doi
- 10.1093/nargab/lqag002
- version
- Version of record, NAR Genomics and Bioinformatics 2026 (PMC12809598 JATS full text)
- retrieved at
- 2026-10-11T21:30:00Z
- artifact sha256
- aae321d05839186a7595987d05af6e95aec5ac32b93b35c5cf1126902fe798b8
- publication status
- peer_reviewed
- licence
- CC BY-NC 4.0
- media type
- application/xml
- venue
- NAR Genomics and Bioinformatics
- year
- 2026
- limitations
- Tables 1 and 2 only: counts of predicted structures per predictive class. The RMSD, LDDT, TM-score, pLDDT and DockQ distributions are figures and supplementary figures and were not extracted.; The counts are class memberships, not a single accuracy figure; direction is recorded as unknown because a high count is good for the accurate classes and bad for the incorrect class.; The ProModEv web portal the paper describes was not retrieved.
- evidence concerns
- source id: lit-structure-20261011-afesmfold-source-mahtha2026; message: The ESMFold rows give 244 EM and 107 NMR monomers, while the AlphaFold2 and AlphaFold3 rows over the same 1666 targets give 161 and 190, and the text gives 160 and 190. The Discussion says ESMFold mispredicted 139 of 190 NMR monomers, against 56 incorrect and 9 special cases in the table. The EM and NMR cells of the ESMFold incorrect row appear transposed.; source locator: Table 1, ESMFold rows, columns 'EM' and 'NMR'; Results paragraph 17; Discussion; artifact sha256: aae321d05839186a7595987d05af6e95aec5ac32b93b35c5cf1126902fe798b8; review date: 2026-10-11; source id: lit-structure-20261011-afesmfold-source-mahtha2026; message: The text says all 11 AlphaFold2 and all 11 AlphaFold3 special cases are NMR structures; the table gives 1 XRD and 10 NMR for AlphaFold2, and 2 XRD, 1 EM and 8 NMR for AlphaFold3.; source locator: Table 1, 'Special cases' rows, columns 'XRD', 'EM', 'NMR'; Results paragraphs 24 and 27; artifact sha256: aae321d05839186a7595987d05af6e95aec5ac32b93b35c5cf1126902fe798b8; review date: 2026-10-11; source id: lit-structure-20261011-afesmfold-source-mahtha2026; message: The row prints EM 3 and NMR 37, so AlphaFold3 totals 184 EM and 49 NMR against 218 and 15 for the other tools over the same 994 targets (text: 219 and 15). The two cells appear transposed.; source locator: Table 2, AlphaFold3 'Acceptable quality' row, columns 'EM' and 'NMR'; artifact sha256: aae321d05839186a7595987d05af6e95aec5ac32b93b35c5cf1126902fe798b8; review date: 2026-10-11; source id: lit-structure-20261011-afesmfold-source-mahtha2026; message: Template quality is a property of the target, yet the column totals differ by tool (TQ 0: 557, 319 and 391). For AlphaFold3 the text gives 414, 326 and 5 accurate and 153, 88 and 8 incorrect dimers by TQ band, against 228, 512 and 4, and 91, 158 and 0, in the table.; source locator: Table 2, columns 'TQ 0', 'TQ 1–40', 'TQ > 40'; Results paragraph 38; artifact sha256: aae321d05839186a7595987d05af6e95aec5ac32b93b35c5cf1126902fe798b8; review date: 2026-10-11; source id: lit-structure-20261011-afesmfold-source-mahtha2026; message: Several rows do not sum to the 'No.' column: Table 1 AlphaFold2 and AlphaFold3 accurate (method columns 1068 and 1075), ESMFold domain movement (429); Table 2 AlphaFold2 incorrect (method 230), AlphaFold3 medium (TQ 305), ESMFold acceptable (TQ 122) and incorrect (method and TQ 584).; source locator: Tables 1 and 2, row totals; artifact sha256: aae321d05839186a7595987d05af6e95aec5ac32b93b35c5cf1126902fe798b8; review date: 2026-10-11
- review
- method: source-hash-verification; method note: Artifact re-retrieved from Europe PMC and SHA-256 matched.; reviewer: claude; reviewer note: Claude review agent reviewer-ps-2, independent of the extractor lit-protein-structure; no human review claimed; date: 2026-10-11