Open-source multi-GPU-accelerated QM/MM simulations with AMBER and QUICK
Using AMBER18 for Relative Free Energy Calculations
Random Forest Refinement of Pairwise Potentials for Protein-Ligand Decoy Detection
The Merz Group is involved in research on several different topics. Some of the projects are briefly described below.
Simulating the Chelate Effect
Comparison of PMF profiles for 12−6 (red) and m12−6−4 (green) Cd2+ion parameters interacting with ethylendiamine.
Despite the rich history of experimental studies focusing on the thermochemistry and kinetics associated with the chelate effect, molecular-level computational studies on the chelate ring opening/ring closure are scarce. The challenge lies in an accurate description of both the metal ion and its aqueous environment. We demonstrated that an optimized 12-6-4 Lennard-Jones (LJ) model can capture the thermodynamics and provide detailed structural and mechanistic insights into the formation of ethylenediamine (en) complexes with metal ions. The water molecules in the first solvation shell of the metal ion were found to facilitate the chelate ring formation. The reported optimized parameters were further able to simulate the formation of bis and tris(en) complexes in solution representing the wide applicability of the 12-6-4 model to simulate coordination chemistry and self-assembly processes.
Protocol used to build up the Random Forest model.
Knowledge-based potentials have generally performed better than physics-based scoring functions in detecting the native structure from a collection of decoy protein structures. Through the use of a reference state, the pure interactions between atom/residue pairs can be obtained through the removal of contributions from ideal-gas state potentials. However, it is a challenge for conventional knowledge-based potentials to assign different importance factors to different atom/residue pairs. In this project we used of the “comparison” concept to generate Random Forest (RF) models that assign different importance factors to atom pair potentials to enhance their ability to identify native proteins from decoy proteins. Individual and combined data sets consisting of 12 decoy sets were used to test the performance of the RF models. We find that RF models increase the recognition of native structures without affecting their ability to identify the best decoy structures. We also created models using scrambled atom types, which create physically unrealistic probability functions in order to test the ability of the RF algorithm to create useful models based on inputted scrambled probability functions. From this test, we found that we were unable to create models that are of similar quality relative to the unscrambled probability functions. We also created uniform probability functions where the peak positions are the same as in the original, but each interaction has the same peak height. Using these uniform potentials, we were able to recover models as good as the ones using the full potentials suggesting all that is important in these models are the experimental peak positions. The KECSA2 potential along with all codes used in this work are available at https://github.com/JunPei000/protein_folding-decoy-set.
Using Ligand Induced Protein Chemical Shift Perturbations to Determine Protein-ligand Structure
Protein chemical shift perturbations (CSPs), upon ligand binding, can be used to refine the structure of a protein-ligand complex by comparing experimental CSPs with calculated CSPs for any given set of structural coordinates. We recently describe a fast and accurate methodology that opens up new opportunities to improve the quality of protein-ligand complexes using nuclear magnetic resonance (NMR) based approaches by focusing on the effect of the ligand on the protein. The new computational approach, 1H empirical chemical shift perturbation (HECSP), rapidly calculates ligand binding induced CSPs for both the backbone and sidechain protons in a protein. Given the dearth of experimental information by which a model could be derived we employed high-quality DFT computations using the automated fragmentation quantum mechanics/molecular mechanics (AF-QM/MM) approach to derive a database of ligand induced CSP’s on a series of protein-ligand complexes. Overall, the empirical HECSP model yielded correlation coefficients between its predicted and DFT computed values of 0.897 (1HA), 0.971 (1HN) and 0.945 (sidechain 1H) with root-mean-square errors (RMSEs) of 0.151 (1HA), 0.199 (1HN) and 0.257 ppm (sidechain 1H), respectively. Using the HECSP model, we also developed a scoring function (NMRScore_P) that can refine complex structures using experimental CSP information. The results of studies on the apo-Neocarzinostatin (apoNCS)-naphthoate and human intestinal fatty acid binding protein (hIFABP)-ketorolac-ANS systems demonstrate that an NMRScore_P strategy for protein-ligand complexes, which is built upon HECSP, can be readily applied to solution NMR structures. In particular, we show the method can distinguish native ligand poses from decoys and refine protein-ligand complex structures. We provide further refined models for both complexes, which satisfy the experimental 1H CSPs. The software to carry out the calculations can be found in the software section (under HECSP) of this web page.
_______________________________________________________________________________________________________________________________
Metal Ion Modeling Using Classical Mechanics
Metal ions play significant roles in numerous fields including chemistry, geochemistry, biochemistry and materials science. With computational tools increasingly becoming important in chemical research, methods have emerged to effectively face the challenge of modeling metal ions in the gas, aqueous and solid phases. Herein we review both quantum and classical modeling strategies for metal ion containing systems that have been developed over the past few decades. This review focuses on classical metal ion modeling based on unpolarized models (including the nonbonded, bonded, cationic dummy atom, and combined models), polarizable models (e.g., the fluctuating charge, Drude oscillator, and the induced dipole models), the angular overlap model, and valence bond based models. Quantum mechanical studies of metal ion containing systems at the semi-empirical, ab initio and density functional levels of theory are reviewed as well with a particular focus on how these methods inform classical modeling efforts. Finally, conclusions and future prospects and directions are offered that will further enhance the classical modeling of metal ion containing systems.
Bringing Clarity to the Prediction of Protein-ligand Binding Free Energies via “blurring”
We described a method to evaluate the free energies of ligand binding utilizing a Monte Carlo estimation of the configuration integrals concomitant with uncertainty quantification. Ensembles for integration are built through systematically perturbing an initial ligand conformation in a rigid binding pocket, which is optimized separately prior to incorporation of the ligand. The procedure producing the ensembles we call “blurring” and is carried out using an in-house developed code. The Boltzmann factor contribution of each pose to the configuration integral is computed and from there the free energy is obtained. Potential function uncertainties are estimated using a fragment-based error propagation method. This method has been applied to a set of small aromatic ligands complexed with T4 Lysozyme L99A mutant. Microstate energies have been determined with the force fields ff99SB and ff94, and the semiempirical method PM6DH2 in conjunction with continuum solvation models including Generalized Born (GB), the Conductor-like Screening Model (COSMO), and SMD. Of the methods studied PM6DH2 based scoring gave binding free energy estimates, which yielded a good correlation to the experimental binding affinities (R2=0.7). All methods overestimated the calculated binding free energies. We trace this to insufficient sampling, the single static protein structure, and inaccuracies in the solvent models we have used in this study.
The Movable Type Method Applied to Protein-ligand Binding
Accurately computing the free energy for biological processes like protein folding or protein-ligand association remains a challenging problem. Both describing the complex intermolecular forces involved and sampling the requisite configuration space make understanding these processes innately difficult. Herein, we address the sampling problem using a novel methodology we term “movable type”. Conceptually it can be understood by analogy with the evolution of printing and, hence, the name movable type. For example, a common approach to the study of protein-ligand complexation involves taking a database of intact drug-like molecules and exhaustively docking them into a binding pocket. This is reminiscent of early woodblock printing where each page had to be laboriously created prior to printing a book. However, printing evolved to an approach where a database of symbols (letters, numerals, etc.) was created and then assembled using a movable type system, which allowed for the creation of all possible combinations of symbols on a given page, thereby, revolutionizing the dissemination of knowledge. Our movable type (MT) method involves the identification of all atom pairs seen in protein-ligand complexes and then creating two databases: one with their associated pairwise distant dependent energies and another associated with the probability of how these pairs can combine in terms of bonds, angles, dihedrals and non-bonded interactions. Combining these two databases coupled with the principles of statistical mechanics allows us to accurately estimate binding free energies as well as the pose of a ligand in a receptor. This method, by its mathematical construction, samples all of configuration space of a selected region (the protein active site here) in one shot without resorting to brute force sampling schemes involving Monte Carlo, genetic algorithms or molecular dynamics simulations making the methodology extremely efficient. Importantly, this method explores the free energy surface eliminating the need to estimate the enthalpy and entropy components individually. Finally, low free energy structures can be obtained via a free energy minimization procedure yielding all low free energy poses on a given free energy surface. Besides revolutionizing the protein-ligand docking and scoring problem this approach can be utilized in a wide range of applications in computational biology which involve the computation of free energies for systems with extensive phase spaces including protein folding, protein-protein docking and protein design.
The Knowledge-based & Empirical Combined Scoring Algorithm (KECSA)
Figure. A protein-ligand structural illustration (using PDBID 1xbc) of how the KECSA statistical potential is modeled. The protein binding site is shown as a grey surface with the ligand located within the binding site surrounded by protein residues which it makes contacts with. The pink dashed lines indicate interactions between certain atom pair types i and j, (i.e. carbonyl oxygen with amine nitrogens in this example) which are defined as "selected interactions" in this manuscript. Green dashed lines indicate all other non-covalent interactions between the protein and ligand atoms in the binding pocket, defined as "background interactions". (a) In the mean force state, the system is filled with all types of interactions. (b) The reference state II contains all the background interactions. (c) Removing all the background interactions from total interactions results in a state with only the selected interactions for each i and j combination.
We developed a novel knowledge-based protein-ligand scoring function that employs a new definition for the reference state, allowing us to relate a statistical potential to a Lennard-Jones (LJ) potential. In this way, the LJ potential parameters were generated from protein-ligand complex structural data contained in the PDB. Forty-nine types of atomic pairwise interactions were derived using this method, which we call the knowledge-based and empirical combined scoring algorithm (KECSA). Two validation benchmarks were used to test the performance of KECSA. The first validation benchmark included two test sets that address the training-set and enthalpy/entropy of KECSA The second validation benchmark suite included two large-scale and five small-scale test sets to compare the reproducibility of KECSA with respect to two empirical score functions previously developed in our laboratory (LISA and LISA+), as well as to other well-known scoring methods. Validation results illustrate that KECSA shows improved performance in all test sets when compared with other scoring methods especially in its ability to minimize the RMSE. LISA and LISA+ displayed similar performance using the correlation coefficient and Kendall τ as the metric of quality for some of the small test sets.
________________________________________________________________________________________________________________________________
Energetics of Zinc-Mediated Interactions in the Allosteric Pathways of Metal Sensor Proteins
A metal-mediated interprotomer hydrogen bond has been implicated in the allosteric mechanism of DNA operator binding in several metal-sensing proteins. Using computational methods, we investigated the energetics of such zinc-mediated interactions in members of the ArsR/SmtB family of proteins (CzrA, SmtB, CadC and NmtR) and the MarR family zinc-uptake repressor AdcR, each of which feature similar interactions, but in sites that differ widely in their allosteric responsiveness. We provided novel structural insight into previously uncharacterized allosteric forms of these proteins using computational methodologies. We find this metal-mediated interaction to be significantly stronger (~8 kcal/mol) at functional allosteric metal binding sites compared to a non-responsive site (CadC) and the apo-proteins. Simulations of the apo-proteins further revealed that the high interaction energy works to overcome the considerable disorder at these hydrogen-bonding sites and functions as a “switch” to lock in a weak DNA-binding conformation once metal is bound. These findings suggested a globally conserved functional role of metal-mediated second-coordination shell hydrogen bonds at allosterically responsive sites in zinc-sensing transcription regulators.
________________________________________________________________________________________________________________________________
Computing ERI's on GPUs
Electron repulsion integral (ERI) calculation on graphical processing units (GPUs) can significantly accelerate quantum chemical calculations. We reported developing an approach to carry out ab initio self-consistent-field (SCF) calculations on GPUs using recurrence relations, which is one of the fastest ERI evaluation algorithms currently available. A direct-SCF scheme to assemble the Fock matrix efficiently was presented, wherein ERIs are evaluated on-the-fly to avoid CPU-GPU data transfer, a well known architectural bottleneck in GPU specific computation. Realized speedups on GPUs reached were from one to two orders of magnitude relative to traditional CPU nodes, with accuracies of better than 1E-7 for systems with more than 4000 basis functions.
________________________________________________________________________________________________________________________________
Wide Open Flaps in Urease
Substrate entry into and product exit from the active site of urease is tightly controlled by an active site flap. Molecular dynamics simulations of urease reveal a previously unobserved, wide-open flap state (see image to the left) that, unlike the well-characterized closed and open states, allows ready access to the metal cluster in the active site. This state is easily reached, via low free energy barriers, from the closed and open states. Additionally, we find that even when the flap is closed, a region of the binding pocket is solvent exposed leading to the hypothesis that it may act as a substrate/product reservoir. The newly identified wide-open state offers further opportunities for small molecule drug discovery by defining a more extensive active site pocket than has been previously described.
________________________________________________________________________________________________________________________________
Metal Ion Recognition by CusF
Escherichia coli CusF is a periplasmic Cu+/Ag+ metallochaperone that provides Cu+/Ag+ efflux by transferring metal ions to the CusCBA tripartite cation efflux pump from the periplasm. The metal recognition site in CusF features a novel cation-p interaction between Cu+/Ag+ and Trp44. Using a variety of computational methodologies; we have provided an exact characterization of the nature and strength of the unique Trp44-Cu+/Ag+ interaction. We quantified the Cu+/Ag+-Trp44 cation-p interaction using high-level ab initio calculations (~14 kcal/mol) performed at the MP2 level of theory. In the absence of structural information, using DFT-QM/MM calculations we determined that Cu+ binds in a distorted tetrahedral coordination geometry in the cation-π interaction lacking Trp44Met mutant from of CusF. In good agreement with experiments, our calculations found the Cu+-Met44 interaction to be stronger than the cation-p interaction in wild type CusF. We have successfully used MD simulations to evaluate the impact of this interaction on the chaperone function of CusF. These simulations suggested an absence of large conformational motions in the protein on the time scale of hundreds of nano-seconds and showed that Trp44 remained pre-organized for metal ion binding in the apo form the protein. We determined that Cu+ binding quenched the protein’s internal motions in residues that participate in the metal-ion dependent interaction with CusB, suggesting protein motions play an essential role in Cu+ transfer to CusB. In addition, our simulations indicate that the strong the cation-p interaction helps maintain the hydrophobic environment around the metal ion and protected it from oxidation. We are currently investigating the nature of the Cu+/Ag+-Trp44 cation-p interactions at the metal recognition in the presence of CusB.
________________________________________________________________________________________________________________________________
Understanding Urea and Urease-Catalyzed Processes
Bosco da Silva
In biological medium urea is catalyzed to the ions ammonium and cyanate by the Ni(II) metalloenzyme urease. A question that one can ask is: How fast is the urease catalyzed reaction? Of course, the answer depends on the uncatalyzed reference reaction in water. To address this question e have used different theoretical approaches to model the mechanism of hydrolysis of urea and thiourea with special focus on understanding the importance of solvent on this process.
Urea is important for the modern (agro)industry; however, urea and other simples species like H2O, NH3, OCNH and CCCO have been observed in the interstellar medium (ISM) and are being proposed to be involved in the formation of relevant biological system like guanine. For these target systems we have proposed new intermediate species for which we have obtained useful descriptors for interstellar detection like the electric dipole moment, rotational constants and the infrared spectrum using high level quantum chemistry calculations. |
What make urease when coming from different sources like animals, plants, fungi and bacteria different? In order to answer this question the 3D electrostatic potential (EP) grid has been calculated for crystallographic structures obtained from different sources. Depending on the size of the urease structure as well as the grid details the number of EP point can easily reach thousands of points. For treating with a such huge data matrix multivariate chemometrics techniques like Principal Component Analysis and Hierarchical Cluster Analysis will be employed.
|
________________________________________________________________________________________________________________________________
Intramolecular Basis Set Superposition Error
We have developed a method for estimating the BSSE of large systems. We use a statistical model to estimate fragment contributions and use error propagation to estimate overall BSSE with uncertainty estimates. This allows for the rapid estimate of BSSE and how it affects the computed electronic energy of a large macromolecule, while at the same time avoiding extra calculations involved in the traditional counterpoise methodology.
MM & QM/MM MD Simulation of Farnesyltransferase
FTase catalyzes the transfer of lipid farnesyl groups required by G protein membrane anchoring and is being studied with MM and QM/MM MD methodologies.
For one substrate An SN2 like reaction mechanism is observed with a ΔGact of ~21 kcal/mol.(experiment: 20.6 kcal/mol), while for a second an associative with dissociative characteristics mechanism is favored. In the case of FTase there is a substrate specific tuning of the preferred nucleophilic substitution reaction mechanism. Mutation studies have also revealed key interaction that stabilizes the transition state.
A similar prenyltransferase enzyme (Orf2) shows a traditional SN1 reaction mechanism because of the nature of its aromatic target, which stabilizes the forming carbocation via cation-pi interactions.
For one substrate An SN2 like reaction mechanism is observed with a ΔGact of ~21 kcal/mol.(experiment: 20.6 kcal/mol), while for a second an associative with dissociative characteristics mechanism is favored. In the case of FTase there is a substrate specific tuning of the preferred nucleophilic substitution reaction mechanism. Mutation studies have also revealed key interaction that stabilizes the transition state.
A similar prenyltransferase enzyme (Orf2) shows a traditional SN1 reaction mechanism because of the nature of its aromatic target, which stabilizes the forming carbocation via cation-pi interactions.
Staphylococcus aureus CzrA
The zinc sensing transcriptional repressor Staphylococcus aureus CzrA is a member of the widely distributed ArsR/SmtB family of prokaryotic metal sensor proteins and provides an excellent model system to understand zinc homeostasis model in cell biology. CzrA is a homodimeric winged helix-turn-helix protein that functions through a mechanism of allosteric negative regulation in which Zn(II) ion binding induces a quaternary structural switch from a “closed” conformation to a more “open” conformation, disrupting the distant DNA binding interface and leading to a low binding affinity for DNA and transcriptional derepression. In close collaboration with the Giedroc group at the University of Indiana, we have used classical MD, QM/MM and QM/MM MD methods to investigate the molecular basis for the large conformational motions and allosteric coupling free energy (~6 kcal/mol) associated with metal ion binding. Our simulations successfully capture the closed to open allosteric switching in DNA bound CzrA on Zn(II) binding. They revealed that zinc binding quenched global conformational sampling by CzrA, whereas DNA binding enhanced the mobility of residues in the allosteric metal binding sites. Our findings are in close agreement with experiments and identified networks of residues involved in correlated and anti-correlated motions that allowed the metal binding and DNA binding sites to communicate. Our analysis of the essential dynamics found metal ion binding to be the primary driving force for the quaternary structural change in CzrA. We also showed that Zn(II) binding limited the conformational space sampled by CzrA, and caused the electrostatic surface potential at the DNA binding interface to become less favorable towards DNA binding. We have provided strong support for a proposed hydrogen-bonding pathway that physically connects the metal binding residue, His97, to the DNA binding interface through the αR helix. This network is present only in the Zn(II)-bound state and our current work focuses on investigating the nature of this hydrogen bonding pathway in CzrA.
Cartoon of thermodynamic cycle depicting the allosteric forms for CzrA. Red represents CzrA while the small green circles are Zn(II) ions. All structures were taken from a snapshot from our MD simulations. The bottom left protein form represents the "closed" form, while the remaining conformations represent the "open" form. Experimentally the vertical equilibria are determined, while computationally we determine the horizontal equilibria.
The image represents the allosteric changes in the transcriptional repressor protein Staphylococcus aureus CzrA upon binding of Zn(II). As Zn(II) binds DNA bound apo-CzrA (upper right image) the DNA binding affinity is reduced by 4 orders of magnitude and the structure of CzrA switches from a “closed” conformation to an “open” conformation facilitating its release from DNA (lower left image).
Mechanistic Studies of Urease
  |
Bound metal ions are essential for the biological function of many proteins. We develop methods to study these proteins using classical and quantum-mechanical models.
A current focus is the nickel-containing enzyme urease (right). In particular we are examine the rol of the active site flap in urease catalysis. |
Calculating Heat of Formation for TM complexes
We are using density functional, MP2 and coupled-cluster methods to calculate the heat of formation for various transition metal containing complexes. Identifying appropriate density functionals for different metal species will allow the selection of accurate and efficient levels of theory for QM regions in QM/MM calculations. No one density functional has been identified as a universal best for the third row transition metals. Using prediction of heat of formation as the selection criterion, the most appropriate functionals for zinc species (cc-pVTZ basis set) are TPSSTPSS and TPSSKCIS. The widely popular B3LYP functional performs quite poorly (cc-pVTZ basis set) by comparison for Zn species.
Estimation of Error in Energy Functions
The ability of computational chemistry to make accurate predictions is highly model-dependent. For very large systems such as proteins, the use of simple parameterized models is necessary due to the high number of degrees of freedom. It has been suspected that errors in such models cancel as system-size and state-sampling increases. Contrarily, we have shown that total error in energy functions increases with system-size. Furthermore, we have shown that overall errors can be estimated and reduced by analyzing computational models through the construction of error probability distribution functions for different classes of molecular interactions. An error correction scheme was developed in which a molecular system is analyzed for the number of each class of fragment-based interactions, followed by the removal of propagated systematic error. The remaining random error can be reported as an error bar. The method has shown benefits for applications including protein-ligand binding, protein folding, and the estimation of intramolecular basis set superposition error. The methodology can also be applied to the optimization of new computational models for macromolecular processes.
|
|
Potential Function Error and Protein Folding
Distortions in computed energy landscapes due to error propagation. If each microstate of a protein under study contains a significant amount of error in its calculated energy (shown here as error bars), computed folding surfaces become distorted with respect to the actual folding surface. This effect introduces difficulty in distinguishing between local minima on the folding surface and in finding the native folds of proteins. This effect is magnified for especially large proteins with many intramolecular contacts contributing to their stable protein folds.
Investigating the Pairwise Additivity of Energy Components
Usually large sizes of biological systems complicate calculation of energies of the systems as a whole which induces efforts of fragmenting those large systems into smaller components and working on the resulting subsystems. This practice arises the question whether the energies of the component systems would sum up to the energy of the entire system. For a model protein-ligand system we have shown that additivity principles are applicable to electronic energies of fragments making up a larger molecular entity. We have employed an energy expansion which decomposes protein-ligand interaction energies into n terms where each term designates the contributions originating from m interacting fragments with m≤n. The outcome of this expansion has been compared to the exact binding energy of the ligand to the protein receptor. The two-body terms have been found to represent a good approximation to the total binding energy of the system, which points to pairwise additivity in the examined case. This basic principle of pairwise additivity is utilized in fragment-based drug design approaches and our results support its continued use. This study can also aid in the validation of non-bonded terms contained within common force fields and in the correction of systematic errors in physics-based score functions.
NMR Refinement
NMR provides both structural and dynamic information on systems in solution.
NMR chemical shifts and chemical shift perturbation of protein and ligand atoms can be accurately calculated with the QM-NMR technique developed in house. It provides invaluable tools to structural refinement and validation, unveiling information that is not readily available from experiments.
NMR chemical shifts and chemical shift perturbation of protein and ligand atoms can be accurately calculated with the QM-NMR technique developed in house. It provides invaluable tools to structural refinement and validation, unveiling information that is not readily available from experiments.
X-Ray Refinement
X-ray crystallography is a major tool in experimental structural biology.
Our QM X-ray refinement method is an innovative technique that is aimed at improving the accuracy and reliability of protein crystal structure determination.
Our QM X-ray refinement method is an innovative technique that is aimed at improving the accuracy and reliability of protein crystal structure determination.
Study of Strain Energy in Protein-bound Drug using X-Ray Refinement
Large coordinate errors will lead to large strain energies in PDB complexes.
Exhaustive sampling of the ligand’s conformational space with high quality energy functions and reliable protein-ligand crystal structures are required for evaluating protein-bound drug strain energy accurately. We took ibuprofen as a test case and there was an 88% reduction in ibuprofen’s conformational strain energy using the QM/MM refined structure versus the original PDB ibuprofen conformations (MP2/CBS calculation result). |
|
Fragment-Based Scoring Function
A central problem in de novo drug design is determining the binding affinity of a ligand with a receptor. We created a new scoring algorithm that estimates the binding affinity of a protein-ligand complex given a three-dimensional structure. The method, LISA (Ligand Identification Scoring Algorithm), uses an empirical scoring function to describe the binding free energy. Interaction terms have been designed to account for van der Waals (VDW) contacts, hydrogen bonding, desolvation effects and metal chelation to model the dissociation equilibrium constants using a linear model. Atom types have been introduced to differentiate the parameters for VDW, H-bonding interactions and metal chelation between different atom pairs. A training set of 492 protein-ligand complexes was selected for the fitting process. Different test sets have been examined to evaluate its ability to predict experimentally measured binding affinities. By comparing with other well known scoring functions, the results show that LISA has advantages over many existing scoring functions in simulating protein-ligand binding affinity, especially metalloprotein-ligand binding affinity.
An improved method (LISA2) is under development, where the potential of mean force (PMF) is to combined with the Lennard-Jones potential for score function parameterization.
An improved method (LISA2) is under development, where the potential of mean force (PMF) is to combined with the Lennard-Jones potential for score function parameterization.
Application of QM and QM/MM to Drug Design Problems
We have applied QM and QM/MM methods to numerous problems related to drug discovery. To achieve this we have developed and are developing high-performance parallel quantum chemistry software at both the semiempirical and ab initio levels of theory. Our ab initio software, QUICK, utilizing graphic processing units with the Tesla(R) platform provided by NVIDIA with the CUDA architecture achieves up to a 37 fold speed up relative to a single CPU node. With the help of GPU chips, ab initio MD is feasible for moderate size system, and it is available within a QM/MM model as well via seamless integration between QUICK and the AMBER software package. We have used the divide-and-conquer method is used to reduce the cost of semiempirical and ab initio calculations and we have extended this to the QM/MM method. Drug design applications include generation of QM based descriptors for QSAR studies and the use of QM and QM/MM methods to score protein ligand complexes.
Inhibiting the Metalloenzyme LpxC
Gram-negative bacteria, including Methicillin-resistant Staphylococcus aureus (MRSA) and bioterror agents such as Yersinia pestis (plague), are potent health threats. New antibiotics are needed to target these bacteria. This project will examine how the key bacterial enzyme LpxC functions and then apply that knowledge towards developing new inhibitors that could lead to novel antibiotic therapies.
We are elucidating the catalytic mechanism using quantum mechanical/molecular mechanical (QM/MM) calculations to analyze the effect of different metal ions and roles that certain key residues play in catalysis. We are also studying protein-ligand interactions and identifying new inhibitors by extending established docking and scoring techniques to metalloenzymes. Additionally, future studies are planned to examine how ligand binding influences protein dynamics using molecular dynamics.
We are elucidating the catalytic mechanism using quantum mechanical/molecular mechanical (QM/MM) calculations to analyze the effect of different metal ions and roles that certain key residues play in catalysis. We are also studying protein-ligand interactions and identifying new inhibitors by extending established docking and scoring techniques to metalloenzymes. Additionally, future studies are planned to examine how ligand binding influences protein dynamics using molecular dynamics.
