Evaluation of the structural models of the human reference proteome: AlphaFold2 versus ESMFold

The human reference proteome is routinely modelled with predictive tools such as AlphaFold2. We recently released a database in which, for each human protein, the AlphaFold2 model is paired with its ESMFold counterpart. The two predictive methods take advantage of different procedures and it is inte...

Full description

Saved in:
Bibliographic Details
Main Authors: Matteo Manfredi, Castrense Savojardo, Pier Luigi Martelli, Rita Casadio
Format: Article
Language:English
Published: Elsevier 2025-06-01
Series:Current Research in Structural Biology
Subjects:
Online Access:http://www.sciencedirect.com/science/article/pii/S2665928X25000042
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850103099111243776
author Matteo Manfredi
Castrense Savojardo
Pier Luigi Martelli
Rita Casadio
author_facet Matteo Manfredi
Castrense Savojardo
Pier Luigi Martelli
Rita Casadio
author_sort Matteo Manfredi
collection DOAJ
description The human reference proteome is routinely modelled with predictive tools such as AlphaFold2. We recently released a database in which, for each human protein, the AlphaFold2 model is paired with its ESMFold counterpart. The two predictive methods take advantage of different procedures and it is interesting to compare them in relation to their quality, particularly when an experimental protein structure is not available. Here, we select three state-of-the-art quality assessment methods and we adopt them to compare 42,942 pairs of models. This procedure helps to find the most reliable models for human proteins, particularly for the set of proteins for which structure prediction methods give dissimilar results. We obtain that when predicted structures are similar, AlphaFold2 models consistently receive higher scores than the ESMFold counterparts. When predicted structures differ, the ESMFold model is the best choice for 49 % of the proteins according to a consensus of the three QA tools.
format Article
id doaj-art-a50bd6f975434d00bf8d8cd18be93be7
institution DOAJ
issn 2665-928X
language English
publishDate 2025-06-01
publisher Elsevier
record_format Article
series Current Research in Structural Biology
spelling doaj-art-a50bd6f975434d00bf8d8cd18be93be72025-08-20T02:39:37ZengElsevierCurrent Research in Structural Biology2665-928X2025-06-01910016710.1016/j.crstbi.2025.100167Evaluation of the structural models of the human reference proteome: AlphaFold2 versus ESMFoldMatteo Manfredi0Castrense Savojardo1Pier Luigi Martelli2Rita Casadio3Biocomputing Group, Dept. of Pharmacy and Biotechnology, University of Bologna, ItalyBiocomputing Group, Dept. of Pharmacy and Biotechnology, University of Bologna, Italy; Corresponding author.Biocomputing Group, Dept. of Pharmacy and Biotechnology, University of Bologna, Italy; Biocomputing Group, AlmaClimate Interdepartmental Center, University of Bologna, Italy; Institute of Biomembrane and Bioenergetics, Italian National Research Council (IBIOM-CNR), ItalyBiocomputing Group, AlmaClimate Interdepartmental Center, University of Bologna, Italy; Institute of Biomembrane and Bioenergetics, Italian National Research Council (IBIOM-CNR), Italy; Corresponding author. Biocomputing Group, AlmaClimate Interdepartmental Center, University of Bologna, Italy.The human reference proteome is routinely modelled with predictive tools such as AlphaFold2. We recently released a database in which, for each human protein, the AlphaFold2 model is paired with its ESMFold counterpart. The two predictive methods take advantage of different procedures and it is interesting to compare them in relation to their quality, particularly when an experimental protein structure is not available. Here, we select three state-of-the-art quality assessment methods and we adopt them to compare 42,942 pairs of models. This procedure helps to find the most reliable models for human proteins, particularly for the set of proteins for which structure prediction methods give dissimilar results. We obtain that when predicted structures are similar, AlphaFold2 models consistently receive higher scores than the ESMFold counterparts. When predicted structures differ, the ESMFold model is the best choice for 49 % of the proteins according to a consensus of the three QA tools.http://www.sciencedirect.com/science/article/pii/S2665928X25000042Human protein structure predictionAI computed structural modelsStructural model quality assessmentModel secondary structureModel solvent accessibilityAlphaFold2
spellingShingle Matteo Manfredi
Castrense Savojardo
Pier Luigi Martelli
Rita Casadio
Evaluation of the structural models of the human reference proteome: AlphaFold2 versus ESMFold
Current Research in Structural Biology
Human protein structure prediction
AI computed structural models
Structural model quality assessment
Model secondary structure
Model solvent accessibility
AlphaFold2
title Evaluation of the structural models of the human reference proteome: AlphaFold2 versus ESMFold
title_full Evaluation of the structural models of the human reference proteome: AlphaFold2 versus ESMFold
title_fullStr Evaluation of the structural models of the human reference proteome: AlphaFold2 versus ESMFold
title_full_unstemmed Evaluation of the structural models of the human reference proteome: AlphaFold2 versus ESMFold
title_short Evaluation of the structural models of the human reference proteome: AlphaFold2 versus ESMFold
title_sort evaluation of the structural models of the human reference proteome alphafold2 versus esmfold
topic Human protein structure prediction
AI computed structural models
Structural model quality assessment
Model secondary structure
Model solvent accessibility
AlphaFold2
url http://www.sciencedirect.com/science/article/pii/S2665928X25000042
work_keys_str_mv AT matteomanfredi evaluationofthestructuralmodelsofthehumanreferenceproteomealphafold2versusesmfold
AT castrensesavojardo evaluationofthestructuralmodelsofthehumanreferenceproteomealphafold2versusesmfold
AT pierluigimartelli evaluationofthestructuralmodelsofthehumanreferenceproteomealphafold2versusesmfold
AT ritacasadio evaluationofthestructuralmodelsofthehumanreferenceproteomealphafold2versusesmfold