EM-Fold: De Novo Folding of α-Helical Proteins Guided by Intermediate-Resolution Electron Microscopy Density Maps

Lindert, Steffen; Staritzbichler, René; Wötzel, Nils; Karakaş, Mert; Stewart, Phoebe L.; Meiler, Jens

doi:10.1016/j.str.2009.06.001

Cited by 78 publications

(116 citation statements)

References 63 publications

(84 reference statements)

Supporting

Mentioning

113

Contrasting

Order By: Relevance

“…However, pioneering studies have demonstrated transmembrane helices as density rods that allowed searching helix layout by Monte Carlo assembly approach. 18 Rosetta, a coarse-grained molecular dynamics approach, was also shown to fit a molecular model into the volume data.…”

Section: Atomic Models and Low-resolution Molecular Imagesmentioning

confidence: 99%

Folding Elastic Transmembrane Helices to Fit in a Low-Resolution Image by Electron Microscopy

Ueno

Kawasaki

Saitô

et al. 2011

J. Bioinform. Comput. Biol.

View full text Add to dashboard Cite

Structure prediction of membrane proteins could be constrained and thereby improved by introducing data of the observed molecular shape. We studied a coarse-grained molecular model that relied on residue-based dummy atoms to fold the transmembrane helices of a protein in the observed molecular shape. Based on the inter-residue potential, the α-helices were folded to contact each other in a simulated annealing protocol to search optimized conformation. Fitting the model into a three-dimensional volume was tested for proteins with known structures and resulted in a fairly reasonable arrangement of helices. In addition, the constraint to the packing transmembrane helix with the twodimensional region was tested and found to work as a very similar folding guide. The obtained models nicely represented α-helices with the desired slight bend. Our structure prediction method for membrane proteins well demonstrated reasonable folding results using a low-resolution structural constraint introduced from recent cell-surface imaging techniques.

show abstract

Section: Atomic Models and Low-resolution Molecular Imagesmentioning

confidence: 99%

Folding Elastic Transmembrane Helices to Fit in a Low-Resolution Image by Electron Microscopy

Ueno

Kawasaki

Saitô

et al. 2011

J. Bioinform. Comput. Biol.

View full text Add to dashboard Cite

show abstract

“…This paper investigates the problem of constructing backbone of a protein when the topology of secondary structures is given. EM-fold uses Rosetta to construct the backbone [20,21]. Pathwalking uses pseudo atoms derived from the density map and a constraint satisfaction solver to place Cα atoms [40].…”

Section: Introductionmentioning

confidence: 99%

“…However, it is still challenging to find a suitable template for many proteins. De novo modeling is an alternative method to derive atomic structures without relying on template structures [19][20][21][22][23][24]. It relies on the detection of secondary structure positions and the connection patterns encoded in the skeleton of the density map.…”

mentioning

confidence: 99%

Deriving Protein Backbone Using Traces Extracted from Density Maps at Medium Resolutions

Nasr

2015

Bioinformatics Research and Applications

View full text Add to dashboard Cite

Abstract. Electron cryomicroscopy is an experimental technique that is capable to produce three dimensional gray-scale images for protein molecules, called density maps. At medium resolution, the atomic details of the molecule cannot be visualized from density maps. However, some features of the molecule can be seen such as the locations of major secondary structures and the skeleton of the molecule. In addition, the order and direction of the detected secondary structure traces can be inferred. We introduce a method to construct the entire model of a protein directly for traces extracted from the density map. The initial results show that this method has good potential. A single model was built for each of the 12 proteins used in the test. The RMSD 100 of the models is slightly improved from our previous method. Keywords:Cryo-EM · Volume image · Skeletonization · Protein modeling · Loop modeling IntroductionElectron cryomicroscopy (cryo-EM) is an emerging technique that produces threedimensional (3D) electron density maps at a wide-range of resolutions [1][2][3][4]. When the resolution of density maps is higher than 4Å, the atomic structure can often be derived [5][6][7][8][9]. At the medium resolutions, such as 5-10Å, the backbone and the characteristic features of amino acids are not resolved. It is still challenging to derive the atomic structure from such a density map. When a component of the protein has atomic structure available, fitting can be performed to derive the atomic structure [10][11][12]. When a homologous model is available, rigid or flexible fitting can be used to derive the atomic structure [13][14][15][16][17][18]. However, it is still challenging to find a suitable template for many proteins. De novo modeling is an alternative method to derive atomic structures without relying on template structures [19][20][21][22][23][24]. It relies on the detection of secondary structure positions and the connection patterns encoded in the skeleton of the density map. A number of computational methods have been developed to detect α-helices from the density maps [25][26][27][28][29][30][31]. Most helices longer than two turns can be detected. Most of the major β-sheets can also be detected using various methods such as SheetTracer, SSEhunter,. By analyzing the twist of β-sheet

show abstract

“…

…”

mentioning

confidence: 99%

“…The models created by EM-Fold for each of the seven successful cases from the benchmark in [1] were subjected to Rosetta loop building and refinement using the new density restraint functionality. Models from the EM-Fold refinement step were taken as start models for the loop construction in Rosetta.…”

mentioning

confidence: 99%

De Novo Atomic-Detail Structure Prediction for Proteins Guided by Medium Resolution Density Maps

2011

View full text Add to dashboard Cite

EM-Fold is a software algorithm that folds proteins into medium resolution density maps obtained by cryo-electron microscopy (cryo-EM) or X-ray crystallography [1]. Models built by EM-Fold and refined by Rosetta generally have root mean square distance deviations (RMSDs) between 4 Å and 7 Å over the full length of the protein. These results demonstrate that it is possible to use computational methods to generate models for proteins that have the correct topology from a medium resolution density map and the primary sequence alone. In short, EMFold is capable to use the experimental data to deduct the topology of the protein by adding helix directionality and connectivity not visibly in the density map. While, side chain conformations in selected helix-helix interfaces were predicted correctly, the models were accurate at atomic detail only in limited regions. Often predicted helices lacked bends or differed in length from helices in the experimental structure. Furthermore, even though the true topology could be enriched by selection of low energy models, it could not be identified by virtue of score alone. On the other hand, the high-resolution experimental structures had considerably better scores than any of the models built using the EM-Fold protocol.It was concluded that model refinement to atomic detail accuracy failed due to insufficient sampling: starting from correct topology models, refinement does not construct models sufficiently close to the native structure to stand out by score -possibly because refinement was not guided by the cryo-EM density map. Therefore larger scale deviations such as length or bending of SSEs cannot be corrected. However, accurate construction of loop regions and side chains depends on models with very high agreement of backbone coordinates within secondary structure elements. The present work demonstrates how atomic-detail not visible in the experimental data can be added when including the electron density map as a restraint in the Rosetta refinement step of . This strength of the Rosetta algorithm was already demonstrated when combined with NMR and EPR experimental data [3][4][5][6][7][8][9].The models created by EM-Fold for each of the seven successful cases from the benchmark in [1] were subjected to Rosetta loop building and refinement using the new density restraint functionality. Models from the EM-Fold refinement step were taken as start models for the loop construction in Rosetta. The loop building (round 1) was followed by two more rounds of identifying the regions of the models that agree least with the density map and then rebuilding these regions and relaxing the entire protein [2]. Table 1 summarizes the results of this work.

show abstract

EM-Fold: De Novo Folding of α-Helical Proteins Guided by Intermediate-Resolution Electron Microscopy Density Maps

Cited by 78 publications

References 63 publications

Folding Elastic Transmembrane Helices to Fit in a Low-Resolution Image by Electron Microscopy

Folding Elastic Transmembrane Helices to Fit in a Low-Resolution Image by Electron Microscopy

Deriving Protein Backbone Using Traces Extracted from Density Maps at Medium Resolutions

De Novo Atomic-Detail Structure Prediction for Proteins Guided by Medium Resolution Density Maps

Contact Info

Product

Resources

About