Multiple Protein Structure Alignment at Scale with FoldMason

bioRxiv (Cold Spring Harbor Laboratory), Год журнала: 2024, Номер unknown

Опубликована: Авг. 1, 2024

Abstract Protein structure is conserved beyond sequence, making multiple structural alignment (MSTA) essential for analyzing distantly related proteins. Computational prediction methods have vastly extended our repository of available proteins structures, requiring fast and accurate MSTA methods. Here, we introduce FoldMason, a progressive method that leverages the alphabet from Foldseek, pairwise aligner, hundreds thousands protein exceeding quality state-of-the-art methods, while two orders magnitudes faster than other FoldMason computes confidence scores, offers interactive visualizations, provides speed accuracy large-scale analysis in era prediction. Using Flaviviridae glycoproteins, demonstrate how FoldMason’s MSTAs support phylogenetic below twilight zone. free open-source software: foldmason.foldseek.com webserver: search.foldseek.com/foldmason .

Язык: Английский

---

BFVD—a large repository of predicted viral protein structures

Nucleic Acids Research, Год журнала: 2024, Номер 53(D1), С. D340 - D347

Опубликована: Ноя. 22, 2024

Abstract The AlphaFold Protein Structure Database (AFDB) is the largest repository of accurately predicted structures with taxonomic labels. Despite providing predictions for over 214 million UniProt entries, AFDB does not cover viral sequences, severely limiting their study. To address this, we created Big Fantastic Virus (BFVD), a 351 242 protein by applying ColabFold to sequence representatives UniRef30 clusters. By utilizing homology searches across two petabases assembled sequencing data, improved 36% these structure beyond ColabFold’s initial results. BFVD holds unique repertoire as 62% its entries show no or low structural similarity existing repositories. We demonstrate how substantial fraction bacteriophage proteins, which remained unannotated based on can be matched similar from BFVD. In that, par AFDB, while holding nearly three orders magnitude fewer structures. an important virus-specific expansion repositories, offering new opportunities advance research. freely downloaded at bfvd.steineggerlab.workers.dev and queried using Foldseek labels bfvd.foldseek.com.

Язык: Английский

---

ModelArchive: a deposition database for computational macromolecular structural models

Journal of Molecular Biology, Год журнала: 2025, Номер unknown, С. 168996 - 168996

Опубликована: Фев. 1, 2025

Язык: Английский

---

AlphaFold-Multimer struggles in predicting PROTAC-mediated protein-protein interfaces

bioRxiv (Cold Spring Harbor Laboratory), Год журнала: 2024, Номер unknown

Опубликована: Март 21, 2024

Abstract In the last few years, predicting structure of PROTAC-mediated complexes emerged as a fundamental step toward design therapeutic modalities for cancer and other conditions. The development AlphaFold2 (AF2) caused paradigm shift in structural biology community. From then onwards, further developments enabled prediction multimeric protein structures while improving calculation efficiency, leading to widespread usage AF2 recent release AF3. However, does not consider ligands, suggesting that ligand-mediated protein-protein interfaces (PPIs) are challenging predict. One main claims AF3 is this limitation has been addressed, but currently released webserver provides only ligands no PROTACs available. article, we benchmark performance on test set 28 dimers, well 326 hetero-dimers orthogonal training set, with special attention interface size presence ligand at interface. We evaluated whether newly model able outperform complexes, despite being include PROTAC prediction. letter, aimed identifying possible reasons why AF-based methods fail predict interfaces. Our results show AF2-multimer predictions sensitive predict, majority models incorrect smallest While it performs reasonably absence ligand, unable reliably. also found significantly improve upon accuracy AF2, still correctly large cases.

Язык: Английский

---

BFVD - a large repository of predicted viral protein structures

bioRxiv (Cold Spring Harbor Laboratory), Год журнала: 2024, Номер unknown

Опубликована: Сен. 9, 2024

The AlphaFold Protein Structure Database (AFDB) is the largest repository of accurately predicted structures with taxonomic labels. Despite providing predictions for over 214 million UniProt entries, AFDB does not cover viral sequences, severely limiting their study. To bridge this gap, we created Big Fantastic Virus (BFVD), a 351,242 protein by applying ColabFold to sequence representatives UniRef30 clusters. BFVD holds unique repertoire as 63% its entries show no or low structural similarity existing repositories. We demonstrate how substantially enhances fraction annotated bacteriophage proteins compared sequence-based annotation using Bakta. In that, on par AFDB, while holding nearly three orders magnitude fewer structures. an important virus-specific expansion structure repositories, offering new opportunities advance research. freely available at https://bfvd.steineggerlab.workers.dev/

Язык: Английский

---

Deciphering bacterial protein functions with innovative computational methods

Trends in Microbiology, Год журнала: 2024, Номер unknown

Опубликована: Дек. 1, 2024

Язык: Английский

---