Despite key advances in our capacity to engineer enzymes, we remain very much in the dark with respect to predicting the effects of mutations on function. A central aspect of our incapacity to predict sequence-function relationships is the fact that proteins are dynamic, yet we rarely treat them as such.
Using the β-lactamase system, we examine the effects of sequence alterations on protein dynamics. Comparison of the CPMG NMR backbone dynamics and molecular dynamics simulations of several β-lactamases reveals them to be unusually rigid, with some motions centered about the active site region. This is consistent with evolutionary conservation of dynamics and suggests a functional role. Sequence changes in the active-site area altered the dynamics, yet catalytic function was maintained. Our results indicate that β-lactamases are highly adaptable. Furthermore, we show that enzyme engineering need not preserve native-like dynamics in order to maintain function.
We then apply computational methods to a cytochrome P450 system, to predict the trajectory of ligand binding and entry into the active-site cavity. We apply the Implicit Ligand Sampling and Adaptive Biasing Force methods to successfully predict, in one single simulation, all residues known to be important for fatty acid substrate binding, thus confirming predictive accuracy. In addition, a new binding residue was identified and experimentally confirmed, and a mechanism for evolutionary protection against CO poisoning is proposed. These new computational biology approaches show great promise to guide efforts to identify functional hotspots for mutation.
Conformational flexibility in structural ensembles is an essential component of enzyme function. While the broad dynamical landscape of proteins is known to promote a number of functional events on multiple time scales, it is yet unknown whether structural and functional enzyme homologs rely on the same concerted residue motions to perform their catalytic function. It has been hypothesized that biologically relevant motions occurring on the millisecond timescale have evolved to promote and/or preserve optimal enzyme catalysis. To verify this hypothesis, we use a combination of NMR relaxation dispersion experiments and computational studies to capture and compare the role of conformational flexibility among members of the glycoside hydrolase (GH) superfamily. We show that millisecond motions can be conserved within similar functional regions of structural homologs. Our studies provide novel information on the evolutionary conservation of conformational flexibility in protein function, further offering additional clues on the mutational tolerance of distinct protein folds in protein engineering applications.
Yeast chorismate mutase is a model system for studying allostery. The enzyme catalyzes the conversion of chorismic acid to prephenic acid, the precursor to phenylalanine and tryptophan. The 60 kDa homodimeric protein adopts a less active T state conformation upon tyrosine binding to the effector site, and a more active R state conformation upon tryptophan binding. Although previous results suggest that the enzyme follows a concerted MWC-type mechanism, our preliminary NMR experiments have suggested that binding of allosteric effectors also changes protein structural dynamics on the microsecond-to-millisecond timescale. The binding of tryptophan results in faster, more pronounced protein dynamics, and the binding of tyrosine tends to quench these motions. NOESY-based experiments are being used to assign 13C-methyl resonances throughout the protein, which will provide a clearer picture of these motions. Altogether, these results suggest that changes in structural dynamics are important even for classical MWC allosteric enzymes, providing additional support for a dynamic model of allostery.
The LacI/GalR family comprises an array of transcriptional repressors involved predominantly in the regulation of metabolic pathways in prokaryotes. Members of the family have an N-terminal DNA-binding domain and a regulatory solute-binding domain that shares strong structural homology with periplasmic-binding proteins. Activation of the DNA-binding domain occurs via allosteric interactions upon effector ligand binding in the solute-binding domain. The aim of this research is to elucidate the molecular basis for the evolution of allostery within the LacI/GalR family by biophysically and functionally characterizing reconstructed ancestral proteins along an evolutionary trajectory. We aim to highlight how this may be applied to engineer novel transcriptional regulators necessary for metabolic engineering and synthetic biology.
We present an information-theoretic MD trajectory analysis approach for comparative dynamics. First, our framework provides a mechanism to compute the information-theoretic free-energy for our choice of the internal coordinates and compare MD trajectories. Second, based on a modified form of KL divergence that we introduce here, we compare the overall dynamics of any local regions of the proteins under consideration. Third, we show how our framework allows us to track the temporal evolution of comparative dynamics. We exemplify our approach by comparing the dynamics of wild-type p53 to its structural and contact point-mutations.
Introducing catalytic function into non-catalytic protein scaffolds remains a major challenge in the field of enzyme design. One explanation why designed enzymes are generally far less effective than their naturally occurring counterparts could be that computational enzyme design approaches often fail to consider the effect of conformational dynamics on enzyme activity. To investigate the role of conformational sampling in the emergence of new catalytic activity, we used ancestral protein reconstruction to characterise the evolutionary trajectory from a non-catalytic solute binding protein (SBP) to a catalytically active cyclohexadienyl dehydratase (CDT). Functional characterisation of extant homologs of CDT and reconstructed ancestral proteins revealed that CDT evolved from a cationic amino acid binding protein via several distinct steps. In addition to the introduction of a reactive catalytic motif within the active site and mutations that improved enzyme-substrate complementarity, remote substitutions were required to refine the structure of the active site and alter the conformational landscape of the enzyme. In particular, X-ray crystal structures and molecular dynamics revealed that substitutions along the evolutionary trajectory of CDT led to the reduction in the sampling of non-catalytic conformational states. This work highlights the importance of protein dynamics in the evolution of effective biological catalysts, and could help guide enzyme design efforts.
Correlation between enzyme function and conformational motions of amino acid networks >10 Å from the active site has been well established for discrete enzyme systems. However, approaches for characterizing dynamical properties across diverse sequence homologs within a family and their correlation with enzyme activity remain challenging. Members of the pancreatic-type ribonuclease (ptRNase) superfamily share similarities in structure and fold, but display large variations in conformational dynamics, catalytic efficiencies, and tissue specific biological activities, making them ideal model systems for probing the relationship between conformational motions and function. As a step towards determining this relationship between dynamics and catalytic efficiency for various members of this broad vertebrate family, we performed the systematic characterization of the intrinsic dynamics of >20 RNases, with experimentally solved structures, over a wide range of time-scales by integrating molecular dynamics simulations and NMR relaxation dispersion experiments. Our results show distinct patterns of dynamical variations between canonical RNases clustered into taxonomic groups, henceforth referred to as subfamilies. We show that conformational motions on the catalytically relevant micro- to milli-second timescale are significantly different for RNases sharing the common fold. Interestingly, sequences sharing similar conformational exchange on this timescale also share similar biological functions. Further, quantitative characterization of pairwise correlations of dynamical properties between the RNase members showed strong correlations within subfamilies that share similar functions. These results suggest that selective pressure for conservation of specific atomic-scale dynamical behaviors, among other factors, may potentially impact distinct biological functions of enzymes sharing the same fold. Further experiments are required to characterize the correlation between conserved dynamical properties and biological function.
We have developed tools to modify enzyme activity and quinary structure in response to chosen external stimuli such as phosphorylation, light and proteolysis. I will describe recent results showing the control of prodrug-activating enzymes using external stimuli for more efficient cancer chemotherapy, and the construction of a phosphorylation-dependent synthetic metabolon (multi-enzyme assembly) for herbicide degradation. The synthetic metabolon was designed to have a hyper-branched self-similar fractal-like structure with a large surface are-to-volume ratio. The dynamic assembly of fractal-like enzyme quinary structure affords a new level of organization of protein structures; implications for the design of dynamic, self-assembling biomaterials will be discussed.
Although the relationship between structure and function in proteins is well established, it is not always adequate to provide a complete understanding of function. Detailed understanding of how conformational dynamics orchestrates function in allosteric regulation of recognition and catalysis at atomic resolution remains ambiguous. We use atomistic molecular dynamics simulations to complement experiments to understand how protein conformational dynamics are coupled to function. We analyze multi-dimensional simulation trajectories by mapping key dynamical features within individual macrostates as residue-residue contacts. In this talk, we will discuss computational studies on members of a ubiquitous family of enzymes that catalyze peptidyl-prolyl bonds and regulate many sub-cellular processes. The effects of substrate binding are observed at locations far beyond the binding sites, implying their importance in allostery. The results provide insights into the general interplay between enzyme conformational dynamics and catalysis from an atomistic perspective that have implications for structure based drug design and protein engineering
Dynamic amino acid interaction networks are important for the function and regulation of tryptophan synthase
Tryptophan synthase (TS) catalyzes the last two steps in the tryptophan biosynthetic pathway. The α and β subunits are connected by a 25Å intramolecular tunnel that channels indole, a product from the α reaction, to the active site in the β subunit. Conformational states of the subunits are highly coordinated making TS an ideal model for understanding substrate channeling and enzyme-enzyme interactions. We have previously used nuclear magnetic resonance chemical shift covariance analysis (CHESCA) to delineate amino acid interaction networks in the α subunit for all states along its catalytic cycle1. Intriguingly, the networks change between the resting state (in the absence of substrates) and the working state (under active catalytic turnover). We have now used 15N R2 relaxation dispersion experiments to map microsecond-to-millisecond conformational events for these states. Our results show that CHESCA-derived network residues are connected by webs of conformationally dynamic residues. These results suggest that there are close connections between network residues and structural dynamics, which likely have consequences for both α subunit catalytic function and interactions with the β subunit2.
1. Axe et al, 2014. J. Am. Chem. Soc. 136, 6818−6821.
2. Axe et al, 2015, Protein Sci. 24, 484–494
It is proposed that internal motions in the enzymes drive the sampling of short-lived minor population of conformations (called as sub-states) that contain features to promote various steps during the function of an enzyme. Therefore, quantitatively characterizing the conformational sub-states in the catalytic cycle of enzymes, including substrate binding, structural rearrangements leading to the transition state, product formation and product release will enable us to obtain detailed insights in the role of protein dynamics in catalysis.
The members of the human pancreatic ribonuclease (RNase) family share common structural scaffold and catalyze the hydrolysis of single stranded ribonucleic acid (ssRNA). However, the catalytic efficiency and rate of dynamics among these enzymes differ by more than a million folds. The member RNases differ in their substrate binding properties and little information is available on their preference for RNA substrates. Therefore, using computer simulations we have modeled binding of Human RNase 1-7 and bovine RNase A with two RNA tetra nucleotide substrates (ACAC and AUAU). Preliminary results indicate diverse binding preferences across the human RNase family. Additionally we have investigated the dynamical events that promote removal of hydrolyzed products. We have sampled the various intermediate stages in the reaction pathway of substrate binding and product removal and are currently working on characterizing the functionally relevant sub-states. Therefore, this study provides insights into the interaction between enzyme and substrate, as well as the dynamical motions related to substrate binding and product removal steps
Enzymes are key players of many important biological processes and understanding their mechanism of action is mandatory for proper pharmaceutical or industrial applications of these macromolecules. In fact, 3D structure, function and dynamics appear to be closely related, recent experimental evidence suggesting that conformational exchange may be involved in promoting catalysis in many enzyme systems, although the mechanisms underlying this atomic flexibility remain unclear. It is still unknown whether sequence and/or structure are evolutionarily conserved to promote flexibility events linked to biological function among protein homologs. In order to tackle these interrogations, we have used NMR to characterize the millisecond timescale conformational exchange in various members of the ribonuclease A superfamily. While these enzymes display very similar structure, their evolutionary distance and diversified biological activities complicate flexibility-function analyses. To solve this issue, we have investigated mammalian homologs of human ribonuclease 3 (Eosinophil Cationic Protein, ECP), comparing the human enzyme with its close ECP homologs from monkeys Macaca fascicularis, Pongo pygmaeus, Pongo abelii and Aotus trivirgatus. Our findings show that conformational exchange in the monkey enzymes strongly resembles that of their human counterpart, with subtle changes in exchange rates and/or localization, thus providing insights into the effects of sequence and phylogenetic diversity on protein dynamics. In parallel, antibacterial assays against E. coli and S. aureus have been performed on these proteins, and we have found that the more the protein sequence diverges from the common ancestor, the more potent its antibacterial activity is. Finally, cytotoxicity of these proteins was evaluated on HeLa cells, and a stark difference was found between human ECP and the monkey enzymes, which were still more potent than RNase A. These experiments are the ground to establish the interdependence that could exist between the functions of these proteins and their atomic flexibility.
In the backdrop of a hostile cellular environment, viral genomes must encode processes to regulate the host cell’s machinery to allow for the efficient replication and packaging of new virus particles. These processes are encoded by relatively small genomes, and so viruses face a formidable information storage problem. Picornaviruses are positive-strand RNA viruses, and first produce a large polyprotein that must then be cleaved to generate functional components. Proteolytic precursors and fully mature forms of viral proteins often have different functions. The picornavirus 3CD protein exemplifies this concept. The 3C protein is a protease and has RNA-binding capabilities, and the 3D protein is an RNA-dependent RNA polymerase (RdRp). The 3CD protein has unique protease and RNA-binding abilities relative to 3C and is devoid of RdRp activity1-3. How these functional changes in 3CD arise is poorly explained by its X-ray crystal structure4, which suggests that the 3C and 3D domains are merely tethered together by a linker (i.e. ‘beads-on-a-string’ type model), and the domains do not interact. Recent computer simulations and small angle X-ray scattering experiments are consistent with a dynamic 3CD conformational ensemble in which transient interactions between the domains may influence function5. Our biophysical studies suggest an even simpler mechanism – extension of the C-terminal tail of 3C by a few amino acid residues dramatically changes function. A network of interactions then transmits these structural/dynamic changes to the protease active site and the RNA-binding site. Proteolytic processing of the C-terminus is then the switch between 3CD and 3C/3D function. Such a simple mechanism may not require any additional domain-domain interactions in the 3CD polyprotein to regulate domain function. In a broader context, we propose that the conformational dynamics of precursor and processed forms of viral proteins regulate the specific function of these molecules during the viral replication cycle to expand the functional proteome beyond the limited storage capacity available within small viral genomes6.
1. Ypma-Wong, M.F., P.G. Dewalt, V.H. Johnson, J.G. Lamb and B.L. Semler, 1988. Protein 3CD is the major poliovirus proteinase responsible for cleavage of the P1 capsid precursor. Virology, 1988, 166, 265-270.
2. Harris, K., S. Reddigari, M. Nicklin, T. Hammerle and E. Wimmer. 1992. Purification and characterization of poliovirus polypeptide 3CD, a proteinase and a precursor for RNA polymerase. J. Virol., 66, 7481-7489.
3. Parsley, T.B., C.T. Cornell and B.L. Semler. 1999. Modulation of the RNA binding and protein processing activities of poliovirus polypeptide 3CD by the viral RNA polymerase domain. J. Biol. Chem., 274, 12867-12876.
4. Marcotte, L.L., A.B. Wass, D.W. Gohara, H.B. Pathak, J.J. Arnold, D.J. Filman, C.E. Cameron and J.M. Hogle. 2007. Crystal structure of poliovirus 3CD protein: virally encoded protease and precursor to the RNA-dependent RNA polymerase. J. Virol., 81, 3583-3596.
5. Moustafa, I.M., D.W. Gohara, A. Uchicha, N. Yennawar and C.E. Cameron. 2015. Conformational ensemble of the poliovirus 3CD precursor observed by MD simulations and confirmed by SAXS: A strategy to expand the viral proteome? Viruses, 7, 5962-5986.
6. Chan, Y.M., I.M. Moustafa, J.J. Arnold, C.E. Cameron and D.D. Boehr. 2016. Long-range communication between different functional sites in the picornaviral 3C protein. Structure, 24, 509-517.
G-Protein coupled receptors (GPCRs) are a large protein family exhibiting 7 transmembrane domains capable of triggering various cellular responses upon binding with a ligand located outside the cell. As such, GPCRs play an important physiological role and are considered as valuable drug targets for the pharmaceutical industry. For many years, GPCRs were considered as on/off switches but studies now tend to demonstrate that their activation is far more complex and occurring through oligomerization and ligand specific conformational changes. The urotensinergic system is composed of a GPCR, namely UT, and 2 endogenous ligands, Urotensin-II (UII) and Urotensin-II related peptide (URP). This system plays a role in cardiovascular homeostasis but albeit the development of antagonists, UT-associated drugs have failed to live up to clinical trial expectations. In an attempt to deepen our knowledge of this system complex pharmacology and to develop specific signaling pathway modulators, we designed ligands based on the premise that UT oligomerization influences intracellular signaling associated to receptor activation. These ligands, named pepducins, are lipidated peptides mimicking UT intracellular loops. Using BRET-based biosensors, we observed that UT-pepducins could modulate Gq, b-arrestin-1, b-arrestin-2 but not G12 activation. Looking at distal effectors, UT-pepducins can trigger with less potency and efficiency ERK1/2 phosphorylation and IP1 production. However, they can modulate cellular proliferation like UII.
G-quadruplexes (GQs) are four-stranded nucleic acid secondary structures formed by guanosine (G)-rich DNA and RNA sequences. It is becoming increasingly clear that cellular processes including gene expression and mRNA translation are regulated by GQs. GQ structures have been extensively characterized, however little attention to date has been paid to their conformational dynamics, despite the fact that many biological GQ sequences populate multiple structures of similar free energies, leading to an ensemble of exchanging conformations. Our lab has developed a suite of methods based on site-directed mutagenesis and thermal denaturation that allows one to dissect GQ conformational ensembles comprising twelve or more distinct states, and to characterize GQ folding pathways. This has allowed us to precisely measure a component of GQ conformational entropy due to sliding of the DNA strands relative to one another. These motions can stabilize GQs, elevating their melting temperatures 10 degrees Celcius or more, with important implications for biological function. We observe strong coupling between the motions of different strands within the same GQ, possibly facilitating allosteric communication between different regions of these molecules. In parallel, an analysis of DNA folding kinetics by thermal hysteresis showed that the rate limiting step for tetrameric GQ assembly changes as a function of temperature, shedding new light on the underlying mechanism. Furthermore, our simple and inexpensive experimental techniques have great promise in further unravelling the complex dynamical behavior nucleic acids.
Macromolecules exist as ensembles of inter-converting structures under physiological conditions. These ensembles are a natural consequence of macromolecular architecture and are critical for macromolecular function. Despite their ubiquity and importance, the consequences of these ensembles for molecular evolution remain largely unknown. Here we show that these molecular ensembles profoundly shape evolution. We generated genotype-phenotype maps using a simple physics-based model of macromolecular ensembles. We found that the ensembles led to epistasis, including high-order epistatic interactions between mutations. Because mutations alter the relative probabilities of all conformations in the ensemble, the quantitative effect of a mutation is different in every genetic background. This causes evolutionary trajectories to become unpredictable as mutations accumulate: the quantitative effect of a mutation early in a trajectory does not predict its effect later. Because these ensembles are a natural aspect of molecular architecture and ubiquitous for function, we expect this is a universal link between the physical underpinnings of biology and the evolutionary process.
Proteins are the molecular machines of life, carrying out complex physical and chemical processes that often require concerted motions of local protein structural elements. Previous efforts to design new proteins for applications in research, industry, and medicine have focused on the creation of sequences that stably adopt a single target structure, ignoring the potential impact of protein dynamics in function. Although computational protein design has enjoyed considerable success in creating new proteins using this approach, most have failed to match the efficiencies that are found in nature because standard methods do not allow for the design of exchange between necessary conformational states on a functionally-relevant timescale. Here, we develop a broadly-applicable computational method to engineer protein dynamics that we term meta-multistate design. We used this methodology to design spontaneous exchange between two novel conformations introduced into the global fold of Streptococcal protein G domain β1. The designed proteins, named DANCERs, for Dynamic And Native Conformational ExchangeRs, are stably folded and exchange between predicted conformational states on the millisecond timescale, as evidenced by nuclear magnetic resonance structures and ZZ-exchange experiments. The successful introduction of defined dynamics on functional timescales paves the way to new applications requiring a protein to spontaneously access multiple conformational states.
Anharmonicity in time-dependent conformational fluctuations is noted to be a key feature
of functional dynamics of biomolecules. While anharmonic events are rare, long timescale (s-ms and beyond) simulations facilitate probing of such events. We have developed anharmonic analysis framework to characterize anharmonic events and enable a deeper analysis of their
functional relevance. This framework provides: (1) a measure for anharmonicity in the
form of higher-order statistics and its variation as a function of time, (2) a story board representation of the simulations to identify key anharmonic conformational events, and (3) identify putative anharmonic conformational substates and visualization of transitions between these substates.
Cytosolic proteins work in cellular environments markedly different from highly dilute aqueous solutions typically used in laboratory experiments. Past investigations have indicated that the surrounding solvent drives protein dynamics and therefore could impact enzyme mechanisms. We have used a combination of stop flow kinetic measurements, X-ray crystallography, quasi-elastic neutron scattering studies and computer simulations to investigate the connection between solvent and enzyme dynamics and its interplay with enzyme mechanism. In particular, the Escherichia coli enzyme dihydrofolate reductase (DHFR) shows that with increasing concentrations of isopropanol in solvent the pH-independent khydride rate decreases more than two fold; the enzyme structure shows no noticeable differences but the dynamical motions are suppressed. More interestingly our investigations shows that the altered motions of DHFR cause significant changes in the enzymeʼs ability to access its functionally relevant conformational sub-states. The enzyme, in the presence of isopropanol, makes less frequent and shorter visits to the conformational sub-states relevant for achieving the transition state, explaining the observed decrease in khydride. Evidence regarding the role of conformational sub-states in enzyme mechanisms from other enzymes (including cyclophilin A and lipase B) will also be presented.
Introduction: Bacterial periplasmic maltose-binding proteins (MBPs) are involved in the transportation of different types of sugars, with varied specificity, through an associated ATP-binding cassette (ABC) transport system. The thermophile Thermotoga maritima was found to have two closely related maltose transport operons (referred to as MalE1 and MalE2), with each containing an MBP. More recently, a third maltose transport operon was identified in T. maritima. Literature indicates a stark difference between substrate recognition and binding by these three MBPs (MalE3), nonetheless there is no clear evidence at the amino acid level that suggests a preferential binding and insights into their binding mechanism. We hypothesize that differences in the structural organization of protein binding site, and the overall dynamics of these proteins help them differentiate between different interacting ligands.
Methods: We employed X-ray crystallography, small-angle X-ray scattering (SAXS), computational modeling and simulations to characterize the structural basis for substrate selection by MalE1-3. A new computational technique, Quasi-anharmonic analysis (QAA), was used to characterize protein conformational sub-states associated with ligand binding.
Results: We performed SAXS on both apo- and ligand-bound proteins and a comparison of the radius of gyration (Rg) score and envelope structure of MBPs suggests a change in the overall conformation of these proteins when they bind to the substrates. We successfully obtained crystals of MalE1, MalE2 and MalE3 in apo and ligand-bound [MalE1 and MalE2 with Maltotetreose (MTT) and MalE3 with maltose (MT)] forms. Additionally, we have collected X-ray diffraction datasets on MalE1-MTT, MalE2-MTT, MalE2apo, and MalE3-MT complexes. We found that the binding site in MalE3 appears small enough to accommodate only a disaccharide, unlike MalE1 and MalE2. We also performed 200 nanoseconds computer simulations to compare the dynamics of these MBPs in presence and in absence of ligands. QAA indicates that the apo structures are fairly flexible and sample multiple sub-states (including the ligand-bound states). Results also indicate a significant reduction in dynamics and conformational sampling of protein once it is bound to the substrate molecule.
Conclusion: Our results support our hypothesis of differential-binding of substrates depending on the size of the binding pocket and the overall dynamics of the protein, which varies significantly between the three MBPs
The talk will focus on the use of NMR spectroscopy to understand how intracellular cyclic nucleotide-dependent conformational switches regulate signaling pathways in eukaryotes. Cyclic nucleotides (cNMPs), such as cAMP and cGMP, bind eukaryotic cNMP-binding domains (CNBs) and control multiple cellular functions (e.g. phosphorylation by protein kinases A and G – PKA and PKG, guanine exchange by the exchange protein directly activated by cAMP – EPAC, and ion channel gating in the hyperpolarization and cyclic-nucleotide modulated ion channels - HCN). Hence, the translational potential arising from the manipulation of cNMP-dependent signaling pathways is high. However, the ubiquity of eukaryotic CNBs also poses a challenge in terms of selectivity. Before the full translational potential of cNMP-signalling can be tapped, it is critical to understand the structural basis for selective cNMP agonism and antagonism, which is being deciphered through recent NMR approaches developed in our laboratory. The comparative NMR analyses of multiple eukaryotic CNBs in PKA, PKG, EPAC and HCN is expected to facilitate the development of selective CNB effectors that may serve as drug leads for the treatment of selected cardiovascular diseases.
We describe our convolutional neural network (CNN) deep learning scoring function for structure-based design and show how we integrate it into an open-source workflow for identifying and targeting allosteric pockets. A CNN scoring function automatically learns the key features of protein-ligand interactions from an explicit, 3D representation of the protein-ligand complex. We train and optimize our CNN scoring functions for pose prediction, virtual screening, and affinity prediction. Additionally, the output of the CNN model can be meaningfully visualized at an atomic granularity.
We provide our latest retrospective evaluations of the performance of the CNN scoring function as well as provide two examples of how we are using our CNN models prospectively to target allosteric sites. Our workflow uses molecular dynamics and fragment docking to identify putative allosteric binding sites. Fragment docking, pharmacophore search, energy minimization and our CNN models are then used to screen for potential inhibitors. The top ranked inhibitors are then simulated to evaluate their allostric effects. The entire workflow can be performed using open-source software, including several software packages and websites developed by us.
Enzymes must be ordered to allow the stabilization of transition states by their active sites, yet dynamic enough to adopt alternative conformations suited to other steps in their catalytic cycles. Here, however, a question arises: how does an enzyme evolve a new function through reorganization the active site architecture while keeping a balance between order and dynamics? To address this question, we generate evolutionary transition of enzymatic functions using laboratory evolution and characterize the process of function transitions using various techniques. In this talk, I present few examples of our laboratory evolution experiments to switch enzymatic activities. I show how changes in protein dynamics are correlated with activity switches, we observed cycles of structural destabilization and repair, evolutionary pressure to 'freeze out' unproductive motions and sampling of distinct conformations with specific catalytic properties in bi-functional intermediates. I also present a new methodology to study the effects of mutations on catalysis, multiple-adaptive landscape analysis, in order to reveal specific interactions between a mutation and parts of the substrate. Finally, I discuss views of protein dynamics, long-range mutational interactions, and enzyme evolution.
Influenza causes hundreds of thousands of deaths each year, and the pandemics that occur roughly four times a century are even more catastrophic. We recently simulated an atomic-resolution model of the entire viral coat, containing 210 million atoms. Ongoing analysis of this simulation provides insights into 1) the impact of structural changes on virulence and 2) novel opportunities for drug development in silico.