Protein structure prediction using sparse NOE and RDC restraints with Rosetta in CASP13

Abstract

Computational methods that produce accurate protein structure models from limited experimental data, for example, from nuclear magnetic resonance (NMR) spectroscopy, hold great potential for biomedical research. The NMR‐assisted modeling challenge in CASP13 provided a blind test to explore the capabilities and limitations of current modeling techniques in leveraging NMR data which had high sparsity, ambiguity, and error rate for protein structure prediction. We describe our approach to predict the structure of these proteins leveraging the Rosetta software suite. Protein structure models were predicted de novo using a two‐stage protocol. First, low‐resolution models were generated with the Rosetta de novo method guided by nonambiguous nuclear Overhauser effect (NOE) contacts and residual dipolar coupling (RDC) restraints. Second, iterative model hybridization and fragment insertion with the Rosetta comparative modeling method was used to refine and regularize models guided by all ambiguous and nonambiguous NOE contacts and RDCs. Nine out of 16 of the Rosetta de novo models had the correct fold (global distance test total score > 45) and in three cases high‐resolution models were achieved (root‐mean‐square deviation < 3.5 å). We also show that a meta‐approach applying iterative Rosetta + NMR refinement on server‐predicted models which employed non‐NMR‐contacts and structural templates leads to substantial improvement in model quality. Integrating these data‐assisted refinement strategies with innovative non‐data‐assisted approaches which became possible in CASP13 such as high precision contact prediction will in the near future enable structure determination for large proteins that are outside of the realm of conventional NMR.

AlphaFold Models of Small Proteins Rival the Accuracy of Solution NMR Structures

“…Remarkably, several CASP13 prediction groups provided models that matched the reference structures and/or fit these NMR data even better than the models generated by conventional expert NMR structure analysis. Notable among these top-performing NMR-guided prediction groups were methods using NMR-guided MELD ( Robertson et al, 2019 ) and NMR-guided Rosetta ( Kuenze and Meiler, 2019 ) methods. Amazingly, some other CASP13 prediction groups provided pure prediction models, which did not use the NMR data at all, that also matched the reference structures better than the models generated with the data by expert data analysis ( Sala et al, 2019 ).…”

Protein Structure Prediction from NMR Hydrogen–Deuterium Exchange Data

“…There have been strides to incorporate experimental techniques with computation, with efforts spanning back to the 1980s with NMR and X-ray crystallography and more recently EPR, MS, and cryo-EM among others. − HDX experiments, originally probed in the 1970s, have been used to map exchange rates onto atomic-resolution structures to assign dynamic properties to otherwise static representations. ,− In the general case, HDX rates have also been coupled to molecular dynamics simulations to explain variation in different regions of a protein. ,− Additionally, these data have been incorporated into protein–protein docking of complexes with known tertiary structure to elucidate quaternary structure. − However, importantly, HDX rates have not yet been used to predict de novo tertiary structure. Previous implementations for structural characterization rely on either homology modeling or some starting structures such as an alternative conformation of a protein or a designed protein. − While there are multiple software packages with impressive results that exist for ab initio structure prediction, such as the co-evolution-dependent neural network AlphaFold, the secondary structure assembling BCL, or iterative threading I-TASSER, none have been coupled to experimental data as frequently or diversely as the Rosetta Modeling Software. ,,,,,− Rosetta ab initio structure prediction allows for the generation of models from amino acid sequence alone, assembling fragments generated from short segments with similar sequences using Monte Carlo sampling combined with a hybrid classical physics and probabilistic knowledge-based scoring function in both coarse-grained and full-atom modeling, similar to other multiscale modeling methods. ,…”

Simultaneous Assignment and Structure Determination of Proteins From Sparsely Labeled NMR Datasets

“…Such an approach simultaneously solves the ambiguity in the dataset and the best structural ensembles compatible with different subsets of the data and the physics model. Other tools like Rosetta have been previously used to predict structures given this type of ambiguous data ( Raman et al, 2010 ; Kuenze and Meiler, 2019 ). We showed during the 13th CASP event that MELDxMD significantly improved the accuracy of the produced structures over other methods.…”