# Highly accurate protein structure prediction with AlphaFold

- **DOI:** [10.1038/s41586-021-03819-2](https://doi.org/10.1038/s41586-021-03819-2)
- **Rating:** Exemplary (4.7 out of 5)
- **Verdict:** AlphaFold delivers a benchmark-validated leap to atomic-scale protein prediction, with shallow alignments as its main constraint.
- **Framework version:** v5.0
- **Pipeline version:** v0.4
- **Evaluation date:** 2026-07-20
- **Evaluation page:** https://www.nabu.science/paper/10.1038%2Fs41586-021-03819-2

## Key claims

1. first computational method that can regularly predict protein structures with atomic accuracy even in cases in which no similar structure is known
   - Support: Well-supported
   - Type: Primary finding
   - Basis: blind CASP14 benchmark shows 0.96 Å median backbone accuracy versus 2.8 Å next best
   - Location: Abstract; Results p. 584; Fig. 1a
2. demonstrating accuracy competitive with experimental structures in a majority of cases and greatly outperforming other methods
   - Support: Well-supported
   - Type: Secondary
   - Basis: CASP14 and all-atom results show large margins over competing entries
   - Location: Abstract; Results p. 584; Fig. 1a
3. the high accuracy that AlphaFold demonstrated in CASP14 extends to a large sample of recently released PDB structures
   - Support: Supported
   - Type: Secondary
   - Basis: post-cutoff PDB validation is broad, with filtering and documented residual failure modes
   - Location: Results p. 584; Fig. 2a; Methods, Test set of recent PDB sequences
4. pLDDT, reliably predicts the Cα local-distance difference test (lDDT-Cα) accuracy of the corresponding prediction
   - Support: Supported
   - Type: Secondary
   - Basis: correlation is substantial at r=0.76, though residual variance remains
   - Location: Results p. 584; Fig. 2c
5. accuracy decreases substantially when the median alignment depth is less than around 30 sequences
   - Support: Well-supported
   - Type: Secondary
   - Basis: explicit limitation is quantified in the MSA-depth analysis
   - Location: MSA depth and cross-chain contacts; Fig. 5a

## Strengths

- **Blind benchmark validates central claim** CASP14 is described as the gold-standard assessment for structure prediction and was unavailable to developers during model training. AlphaFold’s median backbone accuracy of 0.96 Å r.m.s.d.95, compared with 2.8 Å for the next-best method, directly supports the headline performance claim.
  - Dimension: methodological rigour
  - Location: Results p. 584; Fig. 1a
- **Broad post-cutoff validation set** The paper evaluates recent PDB structures deposited after the training data cutoff and applies redundancy and template-similarity filtering. This provides evidence that the CASP14 result is not confined to a single benchmark setting.
  - Dimension: methodological rigour
  - Location: Fig. 2a; Methods, Test set of recent PDB sequences
- **Reproducible release supports deployment** The Code availability section provides source code, trained weights, and inference scripts under an open-source license. The Methods also report software versions, compute requirements, and inference timings, making practical adoption unusually well supported.
  - Dimension: application pathway
  - Location: Code availability; Methods, Training regimen and Inference regimen

## Limitations

- **Accuracy depends on alignment depth** The paper reports that accuracy decreases substantially when median alignment depth falls below around 30 sequences. This limits transferability for targets with sparse evolutionary information.
  - Dimension: application pathway
  - Location: MSA depth and cross-chain contacts; Fig. 5a
- **Complex-interface cases remain harder** AlphaFold is described as weaker for proteins with few intra-chain or homotypic contacts compared with heterotypic contacts. This leaves a meaningful gap for chains whose folded shape is largely determined by interactions in larger complexes.
  - Dimension: methodological rigour
  - Location: MSA depth and cross-chain contacts, p. 588
- **Confidence estimates retain residual error** The pLDDT confidence measure correlates with observed lDDT-Cα accuracy, but the reported Pearson correlation of 0.76 leaves non-trivial residual variance. Users still need caution when relying on confidence scores for difficult downstream decisions.
  - Dimension: methodological rigour
  - Location: Fig. 2c

## Reliability checks

No concerns (4 of 4 checks).

- **Internal coherence:** Clean. The central performance figures are consistent across the abstract, results, and figures, including CASP14 median r.m.s.d.95 and recent PDB validation metrics. Methods, reported outcomes, bootstrap intervals, and confidence-calibration analyses align without detected internal contradictions.
- **Research conduct:** Clean. The study is computational and declares human, animal, and clinical ethics review as not applicable. Patent applications are disclosed, and the Data availability and Code availability sections provide public data sources, trained weights, source code, and inference scripts.
- **Reference integrity:** Clean. All 85 cited references resolved to real publications via PubMed, Crossref, or OpenAlex.
  - References not verified automatically are resolved manually within 24 hours.

## Post-publication record

- **Status:** Clean
- **Summary:** No retraction or expression of concern found in Crossref (includes Retraction Watch).
- **Last checked:** 2026-10-17