2022
DOI: 10.1101/2022.11.21.517405
|View full text |Cite
Preprint
|
Sign up to set email alerts
|

AlphaFold predictions are valuable hypotheses, and accelerate but do not replace experimental structure determination

Abstract: AI-based methods such as AlphaFold have raised the possibility of using predicted models in place of experimentally-determined structures. Here we assess the accuracy of AlphaFold predictions by comparing them to density maps obtained from automated redeterminations of recent crystal structures and to the corresponding deposited models. Some AlphaFold predictions match experimental maps closely, but most differ on a global scale through distortion and domain orientation and on a local scale in backbone and sid… Show more

Help me understand this report
View published versions

Search citation statements

Order By: Relevance

Paper Sections

Select...
2
1
1
1

Citation Types

0
39
0

Year Published

2022
2022
2024
2024

Publication Types

Select...
6
2
1

Relationship

1
8

Authors

Journals

citations
Cited by 41 publications
(39 citation statements)
references
References 43 publications
0
39
0
Order By: Relevance
“…74 However, we found a binding site occlusion that compromises the applicability of the AF2 model for structurebased investigations, as observed also in other studies. 68,80,96 We found that the optimization of the binding site was a necessary step for both HM and AF2 models. The refinement process of the AF2 model was needed not only to improve the performance but also to open the orthosteric binding site and allow the docking of agonists.…”
Section: ■ Conclusionmentioning
confidence: 99%
“…74 However, we found a binding site occlusion that compromises the applicability of the AF2 model for structurebased investigations, as observed also in other studies. 68,80,96 We found that the optimization of the binding site was a necessary step for both HM and AF2 models. The refinement process of the AF2 model was needed not only to improve the performance but also to open the orthosteric binding site and allow the docking of agonists.…”
Section: ■ Conclusionmentioning
confidence: 99%
“…Moreover, we found a binding site occlusion and a scale in the TM7 backbone that can compromise the applicability of the AF model for structure-based investigations, as observed also in other studies. 69,80,97 We found that the optimization of the binding site was a necessary step for both HM and AF2 models. The refinement process of AF2 model was needed not only to improve the performance, but also to open the orthosteric binding site and allow the docking of agonists.…”
Section: Discussionmentioning
confidence: 99%
“…Moreover, we found a binding site occlusion and a scale in the TM7 backbone that can compromise the applicability of the AF model for structure-based investigations, as observed also in other studies. 69, 80, 97…”
Section: Discussionmentioning
confidence: 99%
“…Finally, in a significant number of cases, AlphaFold2 and related approaches do not produce high-confidence structures. It was recently shown that while residues predicted by AlphaFold2 with high confidence (> 90 plDDT) have a very low prediction error (median 0.6 Å), this quickly increases to over 3 Å error for low confidence residues (< 70 plDDT) [162] . For such cases with only low confidence structure information present, we may still have to fall back on sequence-based approaches or utilise embedding techniques as described in Section 3.2 .…”
Section: Computational Representations Of Protein Structuresmentioning
confidence: 99%