Editing Protein structure prediction (section)

===Current AI methods and databases of predicted protein structures===
AlphaFold2, was introduced in CASP14, and is capable of predicting protein structures to near experimental accuracy.<ref name="pmid34265844">{{cite journal |vauthors=Jumper J, Evans R, Pritzel A, Green T, Figurnov M, Ronneberger O, Tunyasuvunakool K, Bates R, Žídek A, Potapenko A, Bridgland A, Meyer C, Kohl SA, Ballard AJ, Cowie A, Romera-Paredes B, Nikolov S, Jain R, Adler J, Back T, Petersen S, Reiman D, Clancy E, Zielinski M, Steinegger M, Pacholska M, Berghammer T, Bodenstein S, Silver D, Vinyals O, Senior AW, Kavukcuoglu K, Kohli P, Hassabis D |display-authors=6| title=Highly accurate protein structure prediction with AlphaFold |journal=Nature |volume=596 |issue=7873 |pages=583–589 |date=August 2021 |pmid=34265844 |pmc=8371605 |doi=10.1038/s41586-021-03819-2|bibcode=2021Natur.596..583J}}</ref> AlphaFold was swiftly followed by RoseTTAFold<ref name="pmid34282049">{{cite journal |vauthors=Baek M, DiMaio F, Anishchenko I, Dauparas J, Ovchinnikov S, Lee GR, Wang J, Cong Q, Kinch LN, Schaeffer RD, Millán C, Park H, Adams C, Glassman CR, DeGiovanni A, Pereira JH, Rodrigues AV, van Dijk AA, Ebrecht AC, Opperman DJ, Sagmeister T, Buhlheller C, Pavkov-Keller T, Rathinaswamy MK, Dalwadi U, Yip CK, Burke JE, Garcia KC, Grishin NV, Adams PD, Read RJ, Baker D|display-authors=6 |title=Accurate prediction of protein structures and interactions using a three-track neural network |journal=Science |volume=373 |issue=6557 |pages=871–876 |date=August 2021 |pmid=34282049 |pmc=7612213 |doi=10.1126/science.abj8754|bibcode=2021Sci...373..871B }}</ref> and later by OmegaFold <!--preprint <ref>https://www.biorxiv.org/content/10.1101/2022.07.21.500999v1 {{bare URL inline|date=December 2022}}</ref> --> and the ESM Metagenomic Atlas.<ref name="pmid36319775">{{cite journal |vauthors=Callaway E |title=AlphaFold's new rival? Meta AI predicts shape of 600 million proteins |journal=Nature |volume=611 |issue=7935 |pages=211–212 |date=November 2022 |pmid=36319775 |doi=10.1038/d41586-022-03539-1 |s2cid=253257926 |doi-access=|bibcode=2022Natur.611..211C }}</ref> 

In a study, Sommer et al. 2022 demonstrated the application of protein structure prediction in genome annotation, specifically in identifying functional protein isoforms using computationally predicted structures, available at https://www.isoform.io.<ref>{{Cite journal |last1=Sommer |first1=Markus J. |last2=Cha |first2=Sooyoung |last3=Varabyou |first3=Ales |last4=Rincon |first4=Natalia |last5=Park |first5=Sukhwan |last6=Minkin |first6=Ilia |last7=Pertea |first7=Mihaela |last8=Steinegger |first8=Martin |last9=Salzberg |first9=Steven L. |date=2022-12-15 |title=Structure-guided isoform identification for the human transcriptome |journal=eLife |volume=11 |language=en |pages=e82556 |doi=10.7554/eLife.82556|pmid=36519529 |pmc=9812405 |doi-access=free}}</ref> This study highlights the promise of protein structure prediction as a genome annotation tool and presents a practical, structure-guided approach that can be used to enhance the annotation of any genome.

In 2024, [[David Baker (biochemist)|David Baker]] and [[Demis Hassabis]] were awarded the [[Nobel Prize in Chemistry]]<ref>{{Cite web |title=Nobel Prize in Chemistry 2024 |url=https://www.nobelprize.org/prizes/chemistry/2024/summary/?utm_source=chatgpt.com |access-date=2025-02-03 |website=NobelPrize.org |language=en-US}}</ref> for their contributions to computational protein modeling, including the development of AlphaFold2, an AI-based model for protein structure prediction. AlphaFold2's accuracy has been evaluated against experimentally determined protein structures using metrics such as [[Root mean square deviation of atomic positions|root-mean-square deviation]] (RMSD).<ref>{{Cite web |title=Computational protein design and protein structure prediction |url=https://www.nobelprize.org/uploads/2024/10/advanced-chemistryprize2024.pdf}}</ref> The median RMSD between different experimental structures of the same protein is approximately 0.6 Å, while the median RMSD between AlphaFold2 predictions and experimental structures is around 1 Å. For regions where AlphaFold2 assigns high confidence, the median RMSD is about 0.6 Å, comparable to the variability observed between different experimental structures. However, in low-confidence regions, the RMSD can exceed 2 Å, indicating greater deviations. In proteins with multiple domains connected by flexible linkers, AlphaFold2 predicts individual domain structures accurately but may assign random relative positions to these domains. Additionally, AlphaFold2 does not account for structural constraints such as the membrane plane, sometimes placing protein domains in positions that would physically clash with the membrane.<ref>{{Cite web |last=EMBL-EBI |title=How accurate are AlphaFold2 structure predictions? {{!}} AlphaFold |url=https://www.ebi.ac.uk/training/online/courses/alphafold/validation-and-impact/how-accurate-are-alphafold-structure-predictions/#:~:text=Analogous%20data%20for%20the%20experimental,less%20reliable%20than%20experimental%20structures. |access-date=2025-02-03 |language=en}}</ref>