Development Of Theoretical Methods For Studying Biological Macromolecules

点击翻译按钮获取中文摘要

Obtaining indirect QM/MM free energies with classical reactive molecular dynamics The use of classical force fields has become a rather indispensable for understanding molecular events. Unfortunately, these studies are primarily restrained to evaluating outcomes in the nonreactive regime, as the breaking and formation of bonds cannot be described with conventional additive force fields. In general, if the system in question changes covalency, a quantum mechanical (QM) energetic description is practically mandated. Unfortunately, the requisite QM/MM simulations for generating the reaction free energy profile are prohibitively expensive. Employing a classical reactive force field, such as the Multi-Surface Adiabatic Reactive Molecular Dynamics (MS-ARMD) approach, can prove a more tenable alternative. Still, the rigorous parameterization is just as challenging when trying to preserve faithfulness to energetic detail and conformational preference. However, by focusing on methods that improve configurational overlap with the desired QM level of theory, it is possible to forgo much of the tedium involved in MS-ARMD's parameterization process. In this scenario, one focuses on using MS-ARMD as a reference potential in the indirect calculation of free energy at the prescribed QM ( or QM/MM) Hamiltonian. The underlying advantage is that imperfections in the classical description can be corrected by equilibrium reweighting or non-equilibrium switching simulations. Two reactions of interest (dissociation of p-nitrophenol phosphate and the Sn2 attack of fluoromethane by chlorine) are underway. Water octanol partition coefficients Water octanol partition coefficient serves as a measure for the lipophilicity of a molecule and is important in the field of drug discovery. A novel method for computational prediction of logarithm of partition coefficient (logP) has been developed using molecular fingerprints and a deep neural network. The machine learning model was trained on a dataset of 12,000 molecules and tested on 2000 molecules. In this article, we present our results for the blind prediction of logP for the SAMPL6 challenge. While the best submission achieved a RMSE of 0.41 logP units, our submission had a RMSE of 0.61 logP units. Overall, we ranked in the top quarter out of the 92 submissions that were made. Our results show that the deep learning model can be used as a fast, accurate and robust method for high throughput prediction of logP of small molecules. A replica exchange umbrella sampling (REUS) method for prediction of binding free energies in SAMPL8 challenge In this study we have designed a new approach to calculate the CB8 host-guest binding free energies in the SAMPL8 blind challenge. Force-matched parameters were developed for each specific host-guest system in multiple level of theory including PM6 and XTB. Here we have performed a replica exchange umbrella sampling (REUS) with a cylindrical potential to each host-guest system for binding free energy calculation. The cylindrical potential restrains the guest molecule along the principal axis of the host and prevents unwanted interactions in the host-guest system for binding free energy calculation. The REUS approach helps to explore a wide range of conformations in the host-guest systems and aids in the convergence of the free energies. The exchange criteria is based on the position of each umbrella with respect to the reaction coordinate. This method was tested on some of the molecules in the SAMPL6 challenge and showed a high similarity with experimental binding affinities. Overall, the REUS method with the cylindrical potential is better suited for host-guest binding free energy calculations due to the higher degree of sampling. QM-MM Modeling: Revisiting of the Double-Link Atom Approach Robust hybrid QM/MM methods are essential for characterizing and determining mechanisms of reactions inside of proteins. Describing the boundary between the quantum and classical regions requires a detailed consideration of the electrostatics in order to produce accurate energetics. Revisiting QM/MM methods developed in-house, we have implemented the double-link atom approach in the CHARMM to make the method readily available for users. Additionally, we have updated the CHARMM interface with open-source (Psi4) and commercial quantum chemistry software packages. To enhance usability and flexibility of the method, force field parameters for amino acids side chains have been reoptimized using both the in the Charmm36 forcefield and the Amber forcefield to balance the inadvertent polarization produced upon severing covalent bonds. The new parameters been tested for the Charmm36 for the reproducibility of physical properties, and the testing for the Amber forcefield parameters are underway. These methods are being used to enhance the characterization of QM cluster models on biocatalysts. Approaches for Predicting Physicochemical Properties Theoretical approaches can be a useful partner for predicting physiochemical properties important in experimental design, from drug discovery and development to studies in toxicology, as such predictive approaches can provide guidance in high-throughput screening for rational design. As a part of the SAMPL6 blind challenge, we considered several tactics towards improving quantum chemical predictions of the partition coefficient for a set of drug-like compounds, by increasing the quality of basis sets, considering tautomerization, and accounting for inhomogeneities in the water and n-octanol phases. Additionally, we employed multiple alchemical approaches to predict the transfer free energies from octanol to water. Lipid Bilayer Quantification Quantitative description of lipid bilayers is critical to understanding their function. The bending modulus, Kc, describes the rigidity with respect to ripple-like deformations and is a key descriptor. Previously derived techniques for obtaining Kc have recently been implemented in a user-friendly tool in our group. Because these techniques require large systems to yield accurate results we have been recently developed and implemented a related strategy, in conjunction with Frank Brown at UCSB, that can be used to study smaller systems. SAMPL8 Host-Guest Binding Challenge and SAMPL7 Membrane Permeability Challenge The accurate calculation of thermodynamic quantities is essential for rational drug design. Arguably, the most important of these is how the free energy changes across various molecular processes. Phenomena such as membrane permeation, ligand binding, and reactivity, are intrinsically governed by the associated free energy difference of a given outcome. In an effort to assess methods for computing free energy differences, as well as quantities of interest, the SAMPL (Statistical Assessment of the Modeling of Proteins and Ligands) competition is a series of challenges where physical properties have to be computed blind on a set of molecules (such as drugs, proteins, ligands), using computational methods. Historically, these challenges have included the identification of major tautomers and protonation states, partition and distribution coefficients, solvation free energies, and the binding affinity of small drug molecules. As part of our commitment to using cutting edge methods and keeping up to date with the computational drug design community as large, we are partaking in two SAMPL challenges: (1) computing binding free energies for drugs of abuse '' (e.g., stimulants, opiates, etc.) to cucurbit8uril and (2) evaluating permeabilities of a collection of molecules across an artificial (PAMPA) membrane. Submissions for the challenges are in progress, and incorporate a mixture of molecular mechanical and quantum chemical approaches.