Session W27: Physics of Proteins: Structure and Dynamics

In vivo protein dynamics

Protein dynamics in cells is subject to both physical and evolutionary constraints. I will discuss a series of experiments and simulations to highlight how protein dynamics is modulated in its native cytoplasmic environment, and how it differs from similar dynamics in aqueous solution. Examples will include the activation of chaperones in the cytoplasm, in-situ protein folding dynamics inside live cells and animals, resolving the dwell times and transition state passage times for protein large-amplitude conformational changes, as well as resolving in-cell dynamics of protein complexes such as microtubules or spliceosomal components.   

Graph Neural Networks as Protein-Protein Interface Scoring Functions

There are numerous applications of machine learning techniques in the biological sciences, including scoring computational models of protein-protein interfaces (PPIs). Graph neural networks (GNNs) have been used to develop deep learning-based PPI scoring functions, since they can map three-dimensional protein structures onto a series of nodes and edges with no loss of information. However, after comparing the performance of current state-of-the-art PPI scoring functions on a large dataset of computational models based on high-resolution x-ray crystal structures of protein-protein heterodimers, we find that the GNN-based scoring functions, GNN-DOVE and Deeprank-GNN-ESM, are outperformed by much simpler physics- and knowledge-based functions. We propose that the lower performance of GNN-based scoring functions stems from the imbalance in the quality of the computational models within the training set. Three of the most frequently used training sets only average 1-10% near-native models per target. We generate a balanced dataset with computational models that are uniformly distributed across the ground truth score, DockQ. We use this balanced dataset to retrain both GNN-DOVE and Deeprank-GNN-ESM and determine the improvement of the scoring accuracy after retraining. In addition, we train a new GNN scoring function on the balanced dataset, using novel architecture and node feature representation to improve current state-of-the-art in PPI scoring functions   

Probing Dynamics of Transferrin and their Interaction with Metal Ions using High Sensitivity Dielectric Spectroscopy

Transferrin is a glycoprotein responsible for ferric-ion delivery to various tissues through blood plasma. The saturation level of the protein plays an essential role in iron metabolism and determining certain diseases such as iron deficiency anemia. Transferrin performs its functions by binding two ferric-ions and an anion, while doing so, the conformation of transferrin would change. To probe the conformational dynamics of transferrin and its interaction with ferric-ion in aqueous solutions at molecular level, we have employed a highly sensitive dielectric megahertz-to-terahertz frequency-domain spectroscopy. The dynamics of the protein strongly depend on the temperature, type of metal ions and protein concentrations. The results have revealed the hydration dynamics and transferrin-ion interactions that determine biochemical functions and reactivity of transferrin. Besides, dielectric spectroscopy being a non-invasive and non-destructive method makes it an excellent candidate as a detection method for transferrin in tissues.   

Modeling Flexible Docking of Protein Monomers using Rigid-body Docking along Minimal Energy Paths

Protein-protein interactions are an essential component of biological function and important targets for protein design. When given two separate proteins that do not undergo conformational changes during binding, rigid-body docking and state-of-the-art scoring methods can be used to accurately predict the coordinates of the docked complex. However, most protein complexes undergo significant conformational changes upon binding, dramatically increasing the complexity of the problem. To better understand this ``flexible" docking process, we first generate the minimal energy path between the x-ray crystal structures of the bound and unbound conformations of the protein heterodimers. We will perform rigid-body docking and model scoring for successive conformations along the minimal energy path, which will allow us to assess the magnitude of the root-mean-square deviations in atomic coordinates beyond which it is difficult to predict protein-protein interfaces using the unbound structures.   

Oral: How to "see" proton transfer during protein function using signature vibrational modes?

Proton transfer is a fundamental process in proteins, underlying bioenergetics, bio-catalysis, and bio-signaling. Understanding the functional mechanism of a protein often demands quantitative structural information on the proton-donor, the proton-acceptor, and the proton-transfer pathway. A main challenge in experimental study of proton transfer during a protein function is lack of effective tools to probe the protonation states of all structural members during proton transfer. The imidazole group of histidine can serve as a proton-donor, proton-acceptor, and a mediator during a proton transfer. We report how to use an integrated approach, combining Fourier transform infrared spectroscopy, histidine specific isotope-editing, site-specific mutation, and density-functional-theory based computational study to develop a general method for probing the three protonation states of histidine. Photoactive yellow protein, a light-activated blue light photoreceptor for bacterial phototaxis, was used as a model system in this study. We will discuss potential applications of this technology.   

Protein cores explore a glassy energy landscape

Over the last 60 years, the general framework that has emerged to describe protein folding is that proteins fold on a funneled energy landscape to their native state at the global energy minimum via an equilibrium process. However, there have been no accurate estimates of the critical folding rate, above which proteins would fold to disordered states. Since the energy minimum for protein cores should correspond to the maximum density, here we develop a purely geometric model of proteins to generate collapsed proteins that are resistant to further compression. First, we find that the packing fraction of protein cores found for all high-quality x-ray crystal structures (φ ~ 0.55) is only obtained in the limit of fast thermal quenching during compression, suggesting that the core packing obtained in protein x-ray crystal structures does not represent a global minimum in energy, but the least dense, yet mechanically stable collapsed state. Furthermore, exploring the energy landscape using slower thermal quenching reveals that core amino acids can possess significantly higher packing fractions than φ ~ 0.55, while satisfying all of the geometric constraints of high-quality protein x-ray crystal structures. We thus provide evidence for the glassy nature of the protein energy landscape, which has important implications for protein structure, dynamics, and function.    

Rotational Effects on Free Energy Landscapes of Diffusing Molecules in Coarse-Grained Models

Accurate coarse-grained modeling of the dynamics of macromolecules with a Generalized Langevin Equation (GLE) requires detailed knowledge of the intramolecular potential of mean force and friction coefficients, which are usually calculated from the analysis of an atomistic simulation.

We demonstrate that the free energy landscape of a diffusing molecule can be distorted by the coupling of internal fluctuations with centrifugal forces. We show how corrections for the rotational entropy affect the potential of mean force and illustrate the formal procedure needed to isolate the correct potential energy surface from the measured one.

Furthermore, we show that these corrections to the modeled potential are, in principle, necessary in order to recover accurate site-specific friction coefficients from the time-lagged correlations of the forces in the GLE. These methods are demonstrated on a toy model of a two-state trimer with exponential memory, showing that the total integral of the memory function and the related friction coefficients can be accurately recovered from an observed trajectory of this system.   

Functionalization for Graphene-based Field Effect Transistor Biosensors

Graphene-based field effect transistors exhibit an excellent ability to detect bio-materials at extremely low concentration. These sensors utilize electronic properties of both graphene and semiconductor interfaces to obtain biodetection with high-sensitivity. In this study, we use hemoglobin protein to couple electrically with graphene and detect electrical signals in these devices. To fabricate these sensors, biochemical functionalization has been employed to ensure a high response from the deposition of the proteins on the surface. This process involves immobilizing anchor molecules to graphene through covalent bonding, then forming amide bonds between the anchors and bioreceptor molecules with high affinity to the bioanalyte. However, molecules with biochemical functionalization also interact with contaminated molecules in the protein solutions, producing unwanted electrical signals. We propose a combined method of electrical and optical techniques, which utilizes absorption properties of the proteins, biochemical functionalization, and optoelectronic properties of graphene field effect transistors to produce a high electrical response when light is illuminated on the surface of the biosensor-bioanalyte system. With this process, we find a high response which is specific to the proteins and depends on both the incident wavelength and concentration of the molecules.   

The minimum and optimal restraints in FRET-assisted protein structural modeling

While our ability to predict protein x-ray crystal structures has improved dramatically via the application of deep learning, much is still unknown about the effects of cellular environments on protein structure, dynamics, and function. Recently, Forster Resonance Energy Transfer (FRET) has been shown to be an effective tool for probing protein structure in vivo. When combined with all-atom molecular dynamics (MD) simulations, these two techniques can dramatically increase our insight into protein structure in vivo. However, given the large number of possible FRET residue pairs to measure, what is the minimum number of FRET pairs needed to determine a conformational change and what is the optimal method for selecting these FRET pairs? Here, we explore state-of-the-art methods for selecting FRET pairs to determine how many pairs N_r are needed to drive a known conformational change between two x-ray crystal structures. We find that it is possible to induce conformational changes using only a small fraction of restraints, N_r/N, where N is the number of amino acids. These results establish the feasibility of FRET-assisted structural modeling and provide a practical approach to planning FRET experiments.    

Investigation of GapR protein-mediated DNA softening and supercoiling using magnetic tweezers

DNA topology plays a crucial role in the cell cycle. One of the most common forms of DNA topology is supercoiling. One approach to determine DNA supercoiling is chromatin immunoprecipitation (ChIP) sequencing of GapR protein, a bacterial protein that preferentially binds overtwisted DNA. Through recent studies, in vivo, GapR was thought to remove positive supercoiling caused by type II topoisomerases during the process of replication and transcription. However, the binding and deformation properties of GapR remain incompletely characterized. Here, we use single-molecule magnetic tweezers to study the binding of GapR to DNA, and its effect on DNA stiffness and twisting. Using the magnetic tweezer on 10kb linearized pFOS1 DNA, we measured the effect of varying GapR concentration on the extension of pFOS1 at different superhelical densities and forces. We found that the GapR protein decreases persistence length and unwinds pFOS1 DNA, shifting the extension versus linking number “hat” curve. We also find that GapR protein supercoils the pFOS1 DNA in a cooperative manner described by the titration curve.   

Molecular docking and dynamics of flexible heteroarotinoids as potential inhibitors against SARS-CoV-2 proteins .

The COVID-19 pandemic caused by the SARS-CoV-2 virus is a great threat to public health due to its high infection and mortality rates. In an effort to find a cure, researchers have been exploring different therapeutics that target the virus proteins with some success. Computational methods have been proven effective in studying biological macromolecule and drug discovery. This study utilizes an in-silico molecular docking approach to screen 26 anti-cancer compounds known as flexible heteroarotinoids against all 24 SARS-CoV-2 proteins. Out of 624 docked complexes analyzed, 69 showed binding energies between -9.0 to -11.6 kcal/mol, indicating good to strong binding affinities, with a binding constant K_D of 100 to 1 nmol. Based on the results, at least five compounds displayed excellent binding affinities against functionally significant proteins in the virus life cycle, such as non-structural protein 2, papain-like protease, non-structural protein 4, proof-reading exoribonuclease, membrane protein, and nucleocapsid protein. Analysis of the structure-activity relationships (SARs) showed that flexible heteroarotinoids with a urea linker instead of a thiourea linker, enhanced hydrophobic side chains in the chromane unit, and CF₃ or OCF₃ functional groups attached to the benzene ring exhibited better binding affinities. Further studies on the dynamics of the results are required to explore the potential of repurposing and developing a potent multi-target drug candidate to combat COVID-19.   

Assessing the current state of computational models for protein-protein interfaces

Computational prediction and design of protein-protein interfaces (PPI) is a difficult task that results in models with a wide range of quality. Thus, PPI scoring algorithms have been designed to classify computational models as high- or low-quality without direct knowledge of the experimentally determined structures. To assess PPI scoring algorithms, we scored models of heterodimeric proteins generated with a rigid-body docking algorithm (ZDOCK) and compared these scores to a measure of structural similarity (i.e. DockQ score) to the associated heterodimeric x-ray crystal structures.  We generated hundreds of thousands of computational models, sampled uniformly over the DockQ score, to create a balanced dataset of redocked models. We scored this dataset using seven state-of-the-art PPI model scoring algorithms. We show that only a small fraction of PPI targets possesses strong correlations (ρ > 0.8) between the model scores and DockQ. We use three physical features to differentiate easy- and hard-to-score targets: interface flatness, fractional interface surface area, and number of interfacial contacts. We show that there are strong correlations between model and DockQ scores for PPI targets with large, intertwined interfaces, yet weak correlations for targets with small, flat interfaces. We create a support vector regression score using these physical features that matches or exceeds the performance of current state-of-the-art PPI scoring functions.