Conventional kinesin (kinesin-1) transports membrane-bounded cargos such as mitochondria and vesicles along microtubules. In vivo it is likely that several kinesins move a single organelle and it is important that they operate in a coordinated fashion so that they do not interfere with each other. Evidence for coordination comes from in vitro assays, which show that the gliding speed of a microtubule driven by many kinesins is as high as one driven by just a single kinesin molecule. Coordination is thought to be facilitated by flexible domains so that when one motor is bound another can work irrespectively of their orientations. The tail of kinesin-1 is predicted to be composed of a coiled-coil with two main breaks, the “swivel” (380-442 Dm numbering) and the hinge (560-624). The rotational Brownian motion of microtubules attached to a glass surface by single kinesin molecules was analyzed and measured the torsion elasticity constant. The deletion of the hinge and subsequent tail domains increase the stiffness of the motor (8±1 kBT/rad) compared to the full length (0.06±0.01 kBT/rad measured previously), but does not impair motor cooperation (700±16nm/s vs. full length 756±55nm/s - speed in high motor density motility assays). Removal of the swivel domain generates a stiff construct (7±1 kBT/rad), which is fully functional at single molecule (657±63nm/s), but it cannot work in large numbers (151±46nm/s). Due to the similar value of flexibility for both short construct (8±11 kBT/rad vs 7±1 1 kBT/rad) and their different behavior at high density (700±16 nm/s vs. 151±46 nm/s) a new hypothesis is presented, the swivel might have a strain dependent conformation. Using Circular Dichroism and Fluorescence the secondary structure of this tail region was studied. The central part of the swivel is dimeric α-helical and it is surrounded by random coils, thereby named helix-coil (HC) region. Furthermore, an experimental set-up is developed to exert a torque on individual kinesin molecules using hydrodynamic flow. The results obtained suggest for the first time the possibility that a structural element within the kinesin tail (HC region) has a force-dependent conformation and that this allows motor cooperation.

The scope of this work focuses on computationally modeling compounds with protein structures. While the impetus of drug discovery is the innovation of new therapeutic molecules, it also involves distinguishing molecules that would not be an effective drug. This can be achieved by inventing new tools or by refining old tools. Virtual screening (VS, also called docking), the computational modeling of a molecule in a receptor structure, is a staple in predicting a molecule's affinity for an intended target. In our Virtual Target Screening system (also called inverse-docking), VS is used to find high-affinity targets, which can potentially explain absorption, distribution, metabolism, and excretion (ADME) of a molecule of interest in the human body. The next project, low-mode docking (LD), attempts to improve VS by incorporating protein flexibility into traditional docking where a static receptor structure has potential to produce poor results due to incorrectly predicted ligand poses. Finally, VS, performed mostly on small molecules, is scaled up to cyclic peptides by employing Monte Carlo simulations and molecular dynamics to mimic the steps of small molecule VS. The first project discussed is Virtual Target Screening (also called inverse-docking) where a small molecule is virtually screened against a library of protein structures. Predicting receptors to which a synthesized compound may bind would give insights to drug repurposing, metabolism, toxicity, and lead optimization. Our protocol calibrates each protein entry with a diverse set of small molecule structures, the NCI Diversity Set I. Our test set, 20 kinase inhibitors, was predicted to have a high percentage of kinase "hits" among approximately 1500 protein structures. Further, approved drugs within the test set generally had better rates of kinase hits. Next, normal mode analysis (NMA), which can computationally describe the fundamental motions of a receptor structure, is utilized to approach the rigid body bias problem in traditional docking techniques. Traditional docking involves the selection of a static receptor structure for VS; however, protein structures are dynamic. Simulation of the induced fit effect in protein-ligand binding events is modeled by full articulation of the approximated large-scale low-frequency normal modes of vibration, or "low-modes," coupled with the docking of a ligand structure. Low-mode dockings of 40 cyclin dependent 2 (CDK2) inhibitors into 54 low-modes of CDK2 yielded minimum root-mean-square deviation (RMSD) values of 1.82 – 1.20 Å when compared to known coordinate data. The choice of pose is currently limited to docking score, however, with ligand pose RMSD values of 3.87 – 2.07 Å. When compared to corresponding traditional dockings with RMSD values of 5.89 – 2.33 Å, low-mode docking was more accurate. The last discussion involves the rational docking of a cyclic peptide to the murine double minute 2 (MDM2) oncoprotein. The affinity for a cyclic peptide (synthesized by Priyesh Jain, McLaughin Lab, University of South Florida), PJ-8-73, in MDM2 was found to be within an order of magnitude of a cyclic peptide from the Robinson Lab at the University of Zurich in Switzerland. Both are Β-hairpin cyclic peptides with IC50 values of 650 nm and 140 nm, respectively. Using the co-crystalized structure of the Robinson peptide (PDB 2AXI), we modeled the McLaughlin peptide based on an important interaction of the 6-chloro-tryptophan residue of the Robinson peptide occupying the same pocket in MDM2 as the tryptophan residue by the native p53 transactivation helical domain. By preserving this interaction in initial cyclic peptide poses, the resulting pose of PJ-8-73 structure in MDM2 possessed comparable active site residue contacts and surface area. These protocols will aid medical research by using computer technology to reduce cost and time. VTS utilizes a unique structural and statistical calibration to virtually assay thousands of protein structures to predict high affinity binding. Determining unintended protein targets aids in creating more effective drugs. In low-mode docking, the accuracy of virtual screening was increased by including the fundamental motions of proteins. This newfound accuracy can decrease false negative results common in virtual screening. Lastly, docking techniques, usually for small molecules, were applied to larger peptide molecules. These modifications allow for the prediction of peptide therapeutics in protein-protein interaction modulation, a growing interest in medicine. Impactful in their own ways, these procedures contribute to the discovery of drugs, whether they are small molecules or cyclic peptides.

Although the tremendous catalytic power of enzymes is widely recognized, their exact mechanisms of action are still a source of debate. In order to elucidate the origin of their power, it is necessary to look at individual residues and atoms, and establish their contribution to ligand binding, activation, and reaction. Given the present limitations of experimental techniques, only computational tools allow for such detailed analysis. During my PhD studies I have applied a variety of computational methods, reviewed in Chapter 2, to the study of two enzymes: DfrB dihydrofolate reductase (DHFR) and methyltetrahydrofolate: corrinoid/iron-sulfur protein methyltransferase (MeTr). ¶ The DfrB enzyme has intrigued microbiologists since it was discovered thirty years ago, because of its simple structure, enzymatic inefficiency, and its insensitivity to trimethoprim. This bacterial enzyme shows neither structural nor sequence similarity with its chromosomal counterpart, despite both catalysing the reduction of dihydrofolate (DHF) using NADPH as a cofactor. As numerous attempts to obtain experimental structures of an enzyme ternary complex have been unsuccessful, I combined docking studies and molecular dynamics simulations to produce a reliable model of the reactive DfrB•DHF•NADPH complex. These results, combined with published empirical data, showed that multiple binding modes of the ligands are possible within DfrB. ¶ Comprehensive sequence and structural analysis provided further insight into the DfrB family. The presence of the dfrB genes within integrons and their level of sequence conservation suggest that they are old structures that had been diverging well before the introduction of trimethoprim. Each monomer of the tetrameric active enzyme presents an SH3-fold domain; this is a eukaryotic auxiliary domain never found before as the sole domain of a protein, let alone as the catalytic one. Overall, DfrB DHFR seems to be a poorly adapted catalyst, a ‘minimalistic’ enzyme that promotes the reaction by facilitating the approach of the ligands rather than by using specific catalytic residues. ¶ MeTr initiates the Wood-Ljungdahl pathway of anaerobic CO2 fixation. It catalyses the transfer of the N5-methyl group from N5-methyltetrahydrofolate (CH3THF) to the cobalt centre of a corrinoid/iron-sulfur protein. For the reaction to occur, the N5 position of CH3THF is expected to be activated by protonation. As experimental studies have led to conflicting suggestions, computational approaches were used to address the activation mechanism. ¶ Initially, I tested the accuracy of quantum mechanical (QM) methods to predict protonation positions and pKas of pterin, folate, and their analogues. Then, different protonation states of CH3THF and active-site aspartic residues were analysed. Fragment QM calculations suggested that the pKa of N5 in CH3THF is likely to increase upon protein binding. Further, ONIOM calculations which accounted for the complete protein structure indicated that active-site aspartic residues are likely to be protonated before the ligand. Finally, solvation and binding free energies of several protonated forms of CH3THF were compared using the thermodynamic integration approach. Taken together, these preliminary results suggest that further work with particular emphasis on the protonation state of active-site aspartic residues is needed in order to elucidate the protonation and activation mechanism of CH3THF within MeTr.

<p>Computational protein design aims at identifying protein mutations and conformations with desired target properties (such as increased protein stability, switch of substrate specificity, or novel function) from a vast combinatorial space of candidate solutions. The development of algorithms to efficiently and accurately solve problems in protein design has thus posed significant computational and modeling challenges. Despite the inherent hardness of protein design, a number of computational techniques have been previously developed and applied to a wide range of protein design problems. In many cases, however, the available computational protein design techniques are deficient both in computational power and modeling accuracy. Typical simplifying modeling assumptions for computational protein design are the rigidity of the protein backbone and the discretization of the protein side-chain conformations. Here, we present the derivation, proofs of correctness and complexity, implementation, and application of novel algorithms for computational protein design that, unlike previous approaches, have provably-accurate guarantees even when backbone or continuous side-chain flexibility are incorporated into the model. We also describe novel divide-and-conquer and dynamic programming algorithms for improved computational efficiency that are shown to result in speed-ups of up to several orders of magnitude as compared to previously-available techniques. Our novel algorithms are further incorporated as part of K*, a provably-accurate ensemble-based algorithm for protein-ligand binding prediction and protein design. The application of our suite of protein design algorithms to a variety of problems, including enzyme redesign and small-molecule inhibitor design, is described. Experimental validation, performed by our collaborators, of a set of our computational predictions confirms the feasibility and usefulness of our novel algorithms for computational protein design.</p> / Dissertation

Conventional kinesin (kinesin-1) transports membrane-bounded cargos such as mitochondria and vesicles along microtubules. In vivo it is likely that several kinesins move a single organelle and it is important that they operate in a coordinated fashion so that they do not interfere with each other. Evidence for coordination comes from in vitro assays, which show that the gliding speed of a microtubule driven by many kinesins is as high as one driven by just a single kinesin molecule. Coordination is thought to be facilitated by flexible domains so that when one motor is bound another can work irrespectively of their orientations. The tail of kinesin-1 is predicted to be composed of a coiled-coil with two main breaks, the “swivel” (380-442 Dm numbering) and the hinge (560-624). The rotational Brownian motion of microtubules attached to a glass surface by single kinesin molecules was analyzed and measured the torsion elasticity constant. The deletion of the hinge and subsequent tail domains increase the stiffness of the motor (8±1 kBT/rad) compared to the full length (0.06±0.01 kBT/rad measured previously), but does not impair motor cooperation (700±16nm/s vs. full length 756±55nm/s - speed in high motor density motility assays). Removal of the swivel domain generates a stiff construct (7±1 kBT/rad), which is fully functional at single molecule (657±63nm/s), but it cannot work in large numbers (151±46nm/s). Due to the similar value of flexibility for both short construct (8±11 kBT/rad vs 7±1 1 kBT/rad) and their different behavior at high density (700±16 nm/s vs. 151±46 nm/s) a new hypothesis is presented, the swivel might have a strain dependent conformation. Using Circular Dichroism and Fluorescence the secondary structure of this tail region was studied. The central part of the swivel is dimeric α-helical and it is surrounded by random coils, thereby named helix-coil (HC) region. Furthermore, an experimental set-up is developed to exert a torque on individual kinesin molecules using hydrodynamic flow. The results obtained suggest for the first time the possibility that a structural element within the kinesin tail (HC region) has a force-dependent conformation and that this allows motor cooperation.

In flexible fitting, the high-resolution crystal structure of a molecule is deformed to optimize its position with respect to a low-resolution density map. Solving the flexible fitting problem entails answering the following questions: (A) How can the crystal structure be deformed? (B) How can the term "optimum" be defined? and (C) How can the optimization problem be solved? In this dissertation, we answer the above questions in reverse order. (C) We develop PFCorr, a non-uniform SO(3)-Fourier-based tool to efficiently conduct rigid-body correlations over arbitrary subsets of the space of rigid-body motions. (B) We develop PF2Fit, a rigid-body fitting tool that provides several useful definitions of the optimal fit between the crystal structure and the density map while using PFCorr to search over the space of rigid-body motions (A) We develop PF3Fit, a flexible fitting tool that deforms the crystal structure with a hierarchical domain-based flexibility model while using PF2Fit to optimize the fit with the density map. Our contributions help us solve the rigid-body and flexible fitting problems in unique and advantageous ways. They also allow us to develop a generalized framework that extends, breadth-wise, to other problems in computational structural biology, including rigid-body and flexible docking, and depth-wise, to the question of interpreting the motions inherent to the crystal structure. Publicly-available implementations of each of the above tools additionally provide a window into the technically diverse fields of applied mathematics, structural biology, and 3D image processing, fields that we attempt, in this dissertation, to span. / text

The selection of the forcefield is a crucial issue in any MD related work and there is no clear indication as to which of the many available forcefields is the best for protein analysis. Many recent literature surveys indicate that MD work may be hindered by two limitations, namely conformational sampling and forcefields used (inaccuracies in the potential energy function may bias the simulation toward incorrect conformations). However, the advances in computing infrastructures, theoretical and computing aspects of MD have paved the way to carry out a sampling on a sufficiently longtime scale, putting a need for the accuracies in the forcefield. Because there are established differences in MD results when using forcefields, we have sought to ask how we could assess common mobility segments from a protein by analysis of trajectories using three forcefields in a similar environment. This is important because, disparate fluctuations appear to be more at flexible regions compared to stiff regions; in particular, flexible regions are more relevant to functional activities of the protein molecule. Therefore, we have tried to assess the similarity in the dynamics using three well-known forcefields ENCAD, CHARMM27 and AMBERFF99SB for 61 monomeric proteins and identify the properties of dynamic residues, which may be important for function. The comparison of popular forcefields with different parameterization philosophy may give hints to improve some of the currently existing agnostics in forcefields and characterization of mobile regions based on dynamics of proteins with diverse folds. These may also give some signature on the proteins at the level of dynamics in relation to function, which can be used in protein engineering studies. Nanosecond level MD simulation(30ns) on 61 monomeric proteins were carried out using CHARMM and AMBER forcefields and the trajectories with ENCAD forcefield obtained from Dynameomics database. The trajectories were first analyzed to check whether structural and dynamic properties from the three forcefields similar choosing few parameters in each case. The gross dynamic properties calculated (root mean square deviation (RMSD), TM-score derived RMSD, radius of gyration and accessible surface area) indicated similarity in many proteins. Flexibility index analysis on 17 proteins, which showed a notable difference in the flexibility, indicated that tertiary interactions (fraction of nonnative stable hydrogen bonds and salt bridges) might be responsible for the difference in the flexibility index. The normalized subspace overlap and shape overlap score taken based on the covariance matrices derived from trajectories indicated that majority of the proteins show a range between 0.3-0.5 indicating that the first principal components from these proteins in different combinations may not match well. These results indicate that although dynamic properties in general are similar in many proteins. However, flexibility index and normalized subspace overlap score indicate that subspaces on the first principal component in many proteins may not match completely. The number of proteins showing a better correlation is higher in CHARMM-AMBER combinations than the other two. The structural features from trajectories have been computed in terms of fraction of secondary structure, hydrogen bonds, salt bridges and native contacts. Although secondary structures and native contacts are well preserved during the simulations, the tertiary interactions (hydrogen bonds) are lost in many proteins and may be responsible for the difference in the some of properties among forcefields. Comparison of simulation results to experimental structures in terms of Root mean square fluctuations, Accessible surface area and radius of gyration indicates that the simulations results are on par with the ones derived from experimental structures. We have tried to assess the flexibility in the proteins using normalized Root mean square fluctuations (nRMSF), which for a residue is the ratio of RMSF from simulation to that of crystal structure. We have selected a threshold for this nRMSF to indicate the mobile regions in a protein based on secondary structure analysis. Based on the threshold of nRMSF and conformational properties (deviation in the dihedral angles), we have classified the residue and evaluated the properties of rigid hinge residues and corresponding mobile residues in terms of residue propensity, secondary structure preference and accessible surface area ranges. Since the rigid dynamic residues represent the inherent mobility, they might be important for function. Therefore, we have tried to assess the functional relevance considering the dynamic mobile residues from each protein from each forcefield simulation with the residues important for the function (taken from literature and databases). It is observed that some residues found to be mobile from the simulation are found to match with the experimental ones, although in many cases the number of these mobile residues is higher compared to the experimental ones. In summary, an analysis of protein simulation trajectories using three forcefields on a set of monomeric protein has shown that the gross structural properties and secondary structures from many proteins remain similar, but there are differences as may be seen from flexibility index. However correlation in parameters from CHARMM and AMBER force field is better compared to other two combinations. The differences seen in some of structural properties may arise mainly due to the loss of few tertiary interactions as indicated by the fraction of native hydrogen bonds and salt bridges. Based on the nRMSF, mobile segments obtained from the simulations were identified, and some of the mobile segments are found to match the functionally important residues from the experimental ones. Our work indicates that there are still some differences in the properties from the simulations, which indicates that care must be exercised when choosing a forcefield, especially assessing the functionally relevant residues from the simulations.