Machine Learning Driven Analysis of Large Scale Simulations Reveals Conformational Characteristics of Ubiquitin Chains.

Understanding the conformational characteristics of protein complexes in solution is crucial for a deeper insight in their biological function. Molecular dynamics simulations performed on high performance computing plants and with modern simulation techniques can be used to obtain large data sets that contain conformational and thermodynamic information about biomolecular systems. While this can in principle give a detailed picture of protein-protein interactions in solution and therefore complement experimental data, it also raises the challenge of processing exceedingly large high-dimensional data sets with several million samples. Here we present a novel method for the characterization of protein-protein interactions, which combines a neural network based dimensionality reduction technique to obtain a two-dimensional representation of the conformational space with a density based clustering algorithm for state detection and a metric which assesses the (dis)similarity between different conformational spaces. This method is highly scalable and therefore makes the analysis of massive data sets computationally tractable. We demonstrate the power of this approach to large scale data analysis by characterizing highly dynamic conformational phase spaces of differently linked ubiquitin (Ub) oligomers from coarse-grained simulations. We are able to extract a protein-protein interaction model for two unlinked Ub proteins which is then used to determine how the Ub-Ub interaction pattern is altered in Ub oligomers by the introduction of a covalent linkage. We find that the Ub chain conformational ensemble depends highly on the linkage type and for some cases also on the Ub chain length. By this, we obtain insight into the conformational characteristics of different Ub chains and how this may contribute to linkage type and chain length specific recognition.

Conformational and functional characterization of artificially conjugated non-canonical ubiquitin dimers.

Ubiquitylation is an eminent posttranslational modification referring to the covalent attachment of single ubiquitin molecules or polyubiquitin chains to a target protein dictating the fate of such labeled polypeptide chains. Here, we have biochemically produced artificially Lys11-, and Lys27-, and Lys63-linked ubiquitin dimers based on click-chemistry generating milligram quantities in high purity. We show that the artificial linkage used for the conjugation of two ubiquitin moieties represents a fully reliable surrogate of the natural isopeptide bond by acquiring highly resolved nuclear magnetic resonance (NMR) spectroscopic data including ligand binding studies. Extensive coarse grained and atomistic molecular dynamics (MD) simulations allow to extract structures representing the ensemble of domain-domain conformations used to verify the experimental data. Advantageously, this methodology does not require individual isotopic labeling of both ubiquitin moieties as NMR data have been acquired on the isotopically labeled proximal moiety and complementary MD simulations have been used to fully interpret the experimental data in terms of domain-domain conformation. This combined approach intertwining NMR spectroscopy with MD simulations makes it possible to describe the conformational space non-canonically Lys11-, and Lys27-linked ubiquitin dimers occupy in a solution averaged ensemble by taking atomically resolved information representing all residues in ubiquitin dimers into account.

EncoderMap(II): Visualizing Important Molecular Motions with Improved Generation of Protein Conformations.

Dimensionality reduction can be used to project high-dimensional molecular data into a simplified, low-dimensional map. One feature of our recently introduced dimensionality reduction technique EncoderMap, which relies on the combination of an autoencoder with multidimensional scaling, is its ability to do the reverse. It is able to generate conformations for any selected points in the low-dimensional map. This transfers the simplified, low-dimensional map back into the high-dimensional conformational space. Although the output is again high-dimensional, certain aspects of the simplification are preserved. The generated conformations only mirror the most dominant conformational differences that determine the positions of conformational states in the low-dimensional map. This allows depicting such differences and-in consequence-visualizing molecular motions and gives a unique perspective on high-dimensional conformational data. In our previous work, protein conformations described in backbone dihedral angle space were used as the input for EncoderMap, and conformations were also generated in this space. For large proteins, however, the generation of conformations is inaccurate with this approach due to the local character of backbone dihedral angles. Here, we present an improved variant of EncoderMap which is able to generate large protein conformations that are accurate in short-range and long-range orders. This is achieved by differentiable reconstruction of Cartesian coordinates from the generated dihedrals, which allows adding a contribution to the cost function that monitors the accuracy of all pairwise distances between the Cα-atoms of the generated conformations. The improved capabilities to generate conformations of large, even multidomain, proteins are demonstrated for two examples: diubiquitin and a part of the Ssa1 Hsp70 yeast chaperone. We show that the improved variant of EncoderMap can nicely visualize motions of protein domains relative to each other but is also able to highlight important conformational changes within the individual domains.

Simulating and analysing configurational landscapes of protein-protein contact formation.

Interacting proteins can form aggregates and protein-protein interfaces with multiple patterns and different stabilities. Using molecular simulation one would like to understand the formation of these aggregates and which of the observed states are relevant for protein function and recognition. To characterize the complex configurational ensemble of protein aggregates, one needs a quantitative measure for the similarity of structures. We present well-suited descriptors that capture the essential features of non-covalent protein contact formation and domain motion. This set of collective variables is used with a nonlinear multi-dimensional scaling-based dimensionality reduction technique to obtain a low-dimensional representation of the configurational landscape of two ubiquitin proteins from coarse-grained simulations. We show that this two-dimensional representation is a powerful basis to identify meaningful states in the ensemble of aggregated structures and to calculate distributions and free energy landscapes for different sets of simulations. By using a measure to quantitatively compare free energy landscapes we can show how the introduction of a covalent bond between two ubiquitin proteins at different positions alters the configurational states of these dimers.

Towards a molecular basis of ubiquitin signaling: A dual-scale simulation study of ubiquitin dimers.

Covalent modification of proteins by ubiquitin or ubiquitin chains is one of the most prevalent post-translational modifications in eukaryotes. Different types of ubiquitin chains are assumed to selectively signal respectively modified proteins for different fates. In support of this hypothesis, structural studies have shown that the eight possible ubiquitin dimers adopt different conformations. However, at least in some cases, these structures cannot sufficiently explain the molecular basis of the selective signaling mechanisms. This indicates that the available structures represent only a few distinct conformations within the entire conformational space adopted by a ubiquitin dimer. Here, molecular simulations on different levels of resolution can complement the structural information. We have combined exhaustive coarse grained and atomistic simulations of all eight possible ubiquitin dimers with a suitable dimensionality reduction technique and a new method to characterize protein-protein interfaces and the conformational landscape of protein conjugates. We found that ubiquitin dimers exhibit characteristic linkage type-dependent properties in solution, such as interface stability and the character of contacts between the subunits, which can be directly correlated with experimentally observed linkage-specific properties.

Author Correction: The structure of the ubiquitin-like modifier FAT10 reveals an alternative targeting mechanism for proteasomal degradation.

The original version of the Supplementary Information associated with this Article inadvertently omitted Supplementary Table 3. The HTML version of the Article has been updated to include a corrected version of the Supplementary Information.

The structure of the ubiquitin-like modifier FAT10 reveals an alternative targeting mechanism for proteasomal degradation.

FAT10 is a ubiquitin-like modifier that directly targets proteins for proteasomal degradation. Here, we report the high-resolution structures of the two individual ubiquitin-like domains (UBD) of FAT10 that are joined by a flexible linker. While the UBDs of FAT10 show the typical ubiquitin-fold, their surfaces are entirely different from each other and from ubiquitin explaining their unique binding specificities. Deletion of the linker abrogates FAT10-conjugation while its mutation blocks auto-FAT10ylation of the FAT10-conjugating enzyme USE1 but not bulk conjugate formation. FAT10- but not ubiquitin-mediated degradation is independent of the segregase VCP/p97 in the presence but not the absence of FAT10's unstructured N-terminal heptapeptide. Stabilization of the FAT10 UBDs strongly decelerates degradation suggesting that the intrinsic instability of FAT10 together with its disordered N-terminus enables the rapid, joint degradation of FAT10 and its substrates without the need for FAT10 de-conjugation and partial substrate unfolding.

Evaluation and Optimization of Interface Force Fields for Water on Gold Surfaces.

The structure and dynamics of water at gold surfaces are important for a variety of applications, including lab on a chip and electrowetting. Classical molecular dynamics (MD) simulations are frequently used to investigate systems with water-gold interfaces, such as biomacromolecules in gold nanoparticle dispersions, but the accuracy of the simulations depends on the suitability of the force field. Density functional theory (DFT) calculations of a water molecule on gold were used as a benchmark to assess force field accuracy. It was found that Lennard-Jones potentials did not reproduce the DFT water-gold configurational energy landscape, whereas the softer Morse and Buckingham potentials allowed for a more accurate representation. MD simulations with different force fields exhibited rather different structural and dynamic properties of water on a gold surface. This emphasizes the need for experimental data and further effort on the validation of a realistic force field for water-gold interactions.