Back

Fold or flop: quality assessment of AlphaFold predictions on whole proteomes

Sarti, E.; Cazals, F.

2026-01-15 bioinformatics
10.64898/2025.12.19.695427 bioRxiv
Show abstract

MotivationReliability of AlphaFold predictions is mainly assessed using the predicted Local Distance Difference Test (pLDDT). For model organisms, 30-40% of residues fall into the low-confidence pLDDT range. Moreover, pLDDT sometimes fails to flag physically implausible structures. This raises two questions: can more robust reliability indicators be identified, and do unreliable predictions share common structural or biophysical features? ResultsWe use packing-based dimensionality reduction and clustering to assess the quality of whole-proteome predictions in the AlphaFold Database (AFDB), tracing connections with domain annotations, intrinsic disorder and stereochamical measures. We thus chart pathological structural motifs and unsupported disorder predictions, that reveal strengths and limitations of current self-assessment metrics and allow us to define a novel, misfold-aware structure quality assessment score. AvailabilityThe code to compute arity maps is available within the Structural Bioinformatics Library. See: AlphaFold analysis, and also Documentation, Applications, Installation guide. The code and data for rerunning analyses are made available at doi.org/10.5281/zenodo.18216693 Contactfrederic.cazals@inria.fr online.

Published in Bioinformatics Advances (predicted rank #7) · training set

Matching journals

The top 3 journals account for 50% of the predicted probability mass.

1
Bioinformatics
◐ 1204 papers in training set
Top 0.6%
30.7%
Similar papers in this journal
2
PLOS Computational Biology
● 1863 papers in training set
Top 3%
12.5%
Similar papers in this journal
3
Nature Communications
○ 5641 papers in training set
Top 21%
7.8%
Similar papers in this journal
50% of probability mass above
4
Journal of Molecular Biology
○ 232 papers in training set
Top 0.3%
6.7%
Similar papers in this journal
5
Nature Methods
○ 385 papers in training set
Top 2%
4.8%
Similar papers in this journal
6
Protein Science
○ 246 papers in training set
Top 1%
4.0%
Bioinformatics Advances · published here
◐ 203 papers in training set
Top 2%
3.2%
8
NAR Genomics and Bioinformatics
◐ 242 papers in training set
Top 2%
2.6%
9
Briefings in Bioinformatics
◐ 354 papers in training set
Top 3%
2.4%
10
BMC Bioinformatics
○ 457 papers in training set
Top 4%
1.7%
11
Scientific Reports
○ 3612 papers in training set
Top 55%
1.7%
12
Proceedings of the National Academy of Sciences
● 2444 papers in training set
Top 32%
1.3%
13
Proteins: Structure, Function, and Bioinformatics
○ 88 papers in training set
Top 1%
1.1%
14
Journal of Structural Biology
○ 64 papers in training set
Top 0.5%
1.1%
15
Journal of Chemical Information and Modeling
● 238 papers in training set
Top 2%
1.1%
16
Peer Community Journal
● 281 papers in training set
Top 4%
1.1%
17
International Journal of Molecular Sciences
○ 494 papers in training set
Top 14%
0.9%
18
Computational and Structural Biotechnology Journal
● 242 papers in training set
Top 6%
0.9%
19
Nature Computational Science
○ 55 papers in training set
Top 2%
0.8%
20
PLOS ONE
● 5266 papers in training set
Top 59%
0.8%
21
Nucleic Acids Research
◐ 1281 papers in training set
Top 13%
0.8%
22
Communications Biology
○ 993 papers in training set
Top 31%
0.8%
23
Molecular Biology and Evolution
◐ 542 papers in training set
Top 6%
0.6%
24
PRX Life
● 42 papers in training set
Top 1%
0.6%
25
Frontiers in Molecular Biosciences
○ 102 papers in training set
Top 3%
0.6%
26
eLife
● 5828 papers in training set
Top 71%
0.6%
27
Biomolecules
○ 100 papers in training set
Top 4%
0.6%

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.