While the vast majority of well-structured single protein chains can now be predicted to high accuracy due to the recent AlphaFold [1] model, the prediction of multi-chain protein complexes remains a challenge in many cases. In this work, we demonstrate that an AlphaFold model trained specifically for multimeric inputs of known stoichiometry, which we call AlphaFold-Multimer, significantly increases accuracy of predicted multimeric interfaces over input-adapted single-chain AlphaFold while maintaining high intra-chain accuracy. On a benchmark dataset of 17 heterodimer proteins without templates (introduced in [2]) we achieve at least medium accuracy (DockQ [3] [≥] 0.49) on 14 targets and high accuracy (DockQ [≥] 0.8) on 6 targets, compared to 9 targets of at least medium accuracy and 4 of high accuracy for the previous state of the art system (an AlphaFold-based system from [2]). We also predict structures for a large dataset of 4,433 recent protein complexes, from which we score all non-redundant interfaces with low template identity. For heteromeric interfaces we successfully predict the interface (DockQ [≥] 0.23) in 67% of cases, and produce high accuracy predictions (DockQ [≥] 0.8) in 23% of cases, an improvement of +25 and +11 percentage points over the flexible linker modification of AlphaFold [4] respectively. For homomeric interfaces we successfully predict the interface in 69% of cases, and produce high accuracy predictions in 34% of cases, an improvement of +5 percentage points in both instances.

/pdf/protein-complex-prediction-with-alphafold-multimer-2ndedz324b.pdf

Protein complex prediction with AlphaFold-Multimer

Highly accurate protein structure predictions by deep neural networks such as AlphaFold2 and RoseTTAFold have tremendous impact on structural biology and beyond. Here, we show that, although these deep learning approaches have originally been developed for the in silico folding of protein monomers, AlphaFold2 also enables quick and accurate modeling of peptide-protein interactions. Our simple implementation of AlphaFold2 generates peptide-protein complex models without requiring multiple sequence alignment information for the peptide partner, and can handle binding-induced conformational changes of the receptor. We explore what AlphaFold2 has memorized and learned, and describe specific examples that highlight differences compared to state-of-the-art peptide docking protocol PIPER-FlexPepDock. These results show that AlphaFold2 holds great promise for providing structural insight into a wide range of peptide-protein complexes, serving as a starting point for the detailed characterization and manipulation of these interactions. 

/pdf/harnessing-protein-folding-neural-networks-for-peptide-w6zo0tnv.pdf

Harnessing protein folding neural networks for peptide–protein docking

Abstract Predicting the structure of interacting protein chains is a fundamental step towards understanding protein function. Unfortunately, no computational method can produce accurate structures of protein complexes. AlphaFold2, has shown unprecedented levels of accuracy in modelling single chain protein structures. Here, we apply AlphaFold2 for the prediction of heterodimeric protein complexes. We find that the AlphaFold2 protocol together with optimised multiple sequence alignments, generate models with acceptable quality (DockQ ≥ 0.23) for 63% of the dimers. From the predicted interfaces we create a simple function to predict the DockQ score which distinguishes acceptable from incorrect models as well as interacting from non-interacting proteins with state-of-art accuracy. We find that, using the predicted DockQ scores, we can identify 51% of all interacting pairs at 1% FPR. 

/pdf/improved-prediction-of-protein-protein-interactions-using-1tuqaybm.pdf

Improved prediction of protein-protein interactions using AlphaFold2

Abstract Most proteins fold into 3D structures that determine how they function and orchestrate the biological processes of the cell. Recent developments in computational methods for protein structure predictions have reached the accuracy of experimentally determined models. Although this has been independently verified, the implementation of these methods across structural-biology applications remains to be tested. Here, we evaluate the use of AlphaFold2 (AF2) predictions in the study of characteristic structural elements; the impact of missense variants; function and ligand binding site predictions; modeling of interactions; and modeling of experimental structural data. For 11 proteomes, an average of 25% additional residues can be confidently modeled when compared with homology modeling, identifying structural features rarely seen in the Protein Data Bank. AF2-based predictions of protein disorder and complexes surpass dedicated tools, and AF2 models can be used across diverse applications equally well compared with experimentally determined structures, when the confidence metrics are critically considered. In summary, we find that these advances are likely to have a transformative impact in structural biology and broader life-science research. 

/pdf/a-structural-biology-community-assessment-of-alphafold2-l5qf0fi1.pdf

A structural biology community assessment of AlphaFold2 applications

Protein structure predictions to atomic accuracy with AlphaFold

Predicting the structure of single-chain proteins is now close to being a solved problem due to the recent achievement of AlphaFold2 (AF2). However, predicting the structure of interacting protein chains is still a challenge. Here, we utilise AF2 to optimise a protocol for predicting the structure of heterodimeric protein complexes using only sequence information. We find that using the default AF2 protocol, 32% of the models in the Dockground test set can be modelled accurately. By tuning the input alignment and identifying the best model, we adjusted the performance to 43%. Our protocol uses MSAs generated by AF2 and MSAs paired on the organism level generated with HHblits. In a more extensive, more realistic, independent test set, the accuracy is 59%. In comparison, the alternative fold-and-dock method RoseTTAFold is only successful in 10% of the cases on this set and traditional docking methods 22%. However, for the traditional method, the performance would be lower if the bound form of both monomers was not known. The success is higher for bacterial protein pairs, pairs with large interaction areas consisting of helices or sheets, and many homologous sequences. We can distinguish acceptable (DockQ>0.23) from incorrect models with an AUC of 0.84 on the test set by analysing the predicted interfaces. At an error rate of 1%, 13% are acceptable (at a 10% error rate, 40% of the models are acceptable). All scripts and tools to run our protocol are freely available at: https://gitlab.com/ElofssonLab/FoldDock.

/pdf/improved-prediction-of-protein-protein-interactions-using-5956nidovu.pdf

Improved prediction of protein-protein interactions using AlphaFold2 and extended multiple-sequence alignments

All cellular functions are governed by complex molecular machines that assemble through protein-protein interactions. Their atomic details are critical to the study of their molecular mechanisms but fewer than 5% of hundreds of thousands of human interactions have been structurally characterized. Here, we test the potential and limitations of recent progress in deep-learning methods using AlphaFold2 to predict structures for 65,484 human interactions. We show that higher confidence models are enriched in interactions supported by affinity or structure based methods and can be orthogonally confirmed by spatial constraints defined by cross-link data. We identify 3,137 high confidence models, of which 1,371 have no homology to a known structure, from which we identify interface residues harbouring disease mutations, suggesting potential mechanisms for pathogenic variants. We find groups of interface phosphorylation sites that show patterns of co-regulation across conditions, suggestive of coordinated tuning of multiple interactions as signalling responses. Finally, we provide examples of how the predicted binary complexes can be used to build larger assemblies. Accurate prediction of protein complexes promises to greatly expand our understanding of the atomic details of human cell biology in health and disease.

https://www.biorxiv.org/content/biorxiv/early/2021/11/09/2021.11.08.467664.full.pdf

Towards a structurally resolved human protein interaction network

Predicting the structure of interacting protein chains is fundamental for understanding the function of proteins. Here, we examine the use of AlphaFold2 (AF2) for predicting the structure of heterodimeric protein complexes. We find that using the default AF2 protocol, 44% of the models in a test set can be predicted accurately. However, by optimising the multiple sequence alignment, we can increase the accuracy to 59%. In comparison, the alternative fold-and-dock method RoseTTAFold is only successful in 10% of the cases on this set, template-based docking 35% and traditional docking methods 22%. We can distinguish acceptable (DockQ>0.23) from incorrect models with an AUC of 0.85 on the test set by analysing the predicted interfaces. The success is higher for bacterial protein pairs, pairs with large interaction areas consisting of helices or sheets, and many homologous sequences. Further, we test the possibility to distinguish interacting from non-interacting proteins and find that by analysing the predicted interfaces, we can separate truly interacting from non-interacting proteins with an AUC of 0.82 in the ROC curve, compared to 0.76 with a recently published method. In addition, when using a more realistic negative set, including mammalian proteins, the identification rate remains (AUC=0.83), resulting in that 27% of interactions can be identified at a 1% FPR. All scripts and tools to run our protocol are freely available at: https://gitlab.com/ElofssonLab/FoldDock.

/pdf/improved-prediction-of-protein-protein-interactions-using-30t5mdjc6r.pdf

Gabriele Pozzati

Papers

Improved prediction of protein-protein interactions using AlphaFold2 and extended multiple-sequence alignments

Towards a structurally resolved human protein interaction network

Improved prediction of protein-protein interactions using AlphaFold2