Preface
My research interest includes methodological advances of chemical phenomena. I want to use my expertise of computing, quantum chemistry and molecular mechanics to improve the accuracy and precision of current drug and catalyst design protocols. This involves formulation, development, implementation and application of QM mediated free energy estimation methodologies as well as ease of use of the entire computational exercise. The scope of this work involves use of a range of chemical models with increasing complexity such as molecular docking, molecular mechanics and quantum mechanics. I have worked on each of these components during my research career as PhD and Postdoc. Following is brief description of key areas which I have been worked in.
Development of Machine learning model to train charges and polarizabilities
CHARMM Drude force-field uses charged auxiliary (Drude) particles attached to the non-hydrogen atoms via a harmonic spring to address electronic degree of freedom. Development of automated tool to predict electrostatic parameters of Drude polarizable FFs is quite challenging, as it requires a balance of partial atomic charges, atomic polarizabilities, Thole scale factors, and treatment of lone pairs. We developed a machine learning model to estimate QM-level RESP atomic charges, atomic polarizability and Thole scale factors (which has no QM analog). The approach includes the calculation of atomic polarizabilities and partial atomic charges of about 50K molecules (including 900 FDA approved drugs) using QM-based methods to produce a large training-set, which is be used to build and optimize the machine learning model for rapid parameter estimation.
Parameterization of CHARMM force-field
|
CHARMM force-field describes the potential energy of a molecule in terms of bond, angle, dihedral, improper, electrostatic and van der Waals parameters. This force-field can be broadly categorized into Additive and Drude FF, the latter of which includes polarizabilities along with atomic charges as a part of electrostatic parameters. It contains carefully optimized parameters for biological systems, including nucleic acids, proteins, lipids etc., such that both microscopic and macroscopic properties are well-reproduced. However, parametrization of a new molecular system from first principles is a tedious and exhaustive process which usually requires lot of scripts and programs. To simplify the process, we developed a standalone python package FFParam which provides one platform to assist various aspects of parameterization process. It can handle both CHARMM Additive and Drude parameterization process wherein it can create, run, extract and analyze QM data from Gaussian and Psi4 and MM data from CHARMM and OpenMM. FFParam has cross-platform compatible GUI which assists display of text, plot, and 3D- molecular geometry.
|
Drug-design using SILCS approach
SILCS is a molecular docking technique developed by Alexander D. MacKerell Jr. and group. It combines molecular dynamics and Monte Carlo techniques to calculate grid free energy of a range of chemically diverse ligands. SILCS approach results into SILCS FragMaps representative of LGFE of basic ligand types. These FragMaps can then be used for designing drugs which show maximum affinity of binding. This is an ongoing project, the details of which will soon be released on the website.
Development of CHARMM-PSI4 Interface for faster QMMM calculation
CHARMM simulation program has been interfaced with various QM packages like Gaussian, QChem and Turbomole for performing QMMM calculations. Despite CHARMM being available free to academic users, all the integrated QM programs require external licenses. We thus integrated Psi4 package, which is a open source QM package, and also provides substantial reduction in computational time which is a major bottleneck in any QMMM calculation. The IO between the two programs is efficiently integrated using json format. The electrostatic embedding has been successfully implemented and polarized embedding with Drude FF is ongoing. The CHARMM-Psi4 interface is also being extended to solve QMMM free energy computations.
Lone pairs: An Electrostatic Perspective
|
Using the topological characteristics of molecular electrostatic potential (MESP), we offered a clear-cut quantitative definition of lone pairs. The largest eigenvalue and corresponding eigenvector of the Hessian at the minima are shown to distinguish lone pair regions from the other types of electron localization (such as π bonds). A comparative study of lone pairs as depicted by various other scalar fields such as the Laplacian of electron density and electron localization function is made. Further, an attempt has been made to generalize the definition of lone pairs to the case of cations.
Our article is used as reference to define lone pair by Wikipedia. en.wikipedia.org/wiki/Lone_pair |
Electric Field of molecules
|
The zero-flux surface (ZFS) of the electric field of the molecules introduces the concept of Atoms in Molecules (AIM) based on the field of molecular electrostatic potential (MESP). ZFS manifests the MESP-based atomic basins and consequently brings out the asymmetric electronic distribution of the molecule. For example, an electron-rich atom among the two bonded atoms possesses a completely closed MESP-based atomic basin. Electronic behaviour of molecules such as BF, BH3, AlCl3, B2H6, and Al2Cl6, and a Lewis acid–base pair, viz. NH3BH3, that cannot be explained by classical definition of electronegativity difference can be clearly illustrated through the nature of atomic basins. MESP-based atomic basins can also be used to explain the intricate charge transfer in transition metal complexes, viz. Ni(CO)4, Fe(CO)5, Cr(CO)6, Mn2(CO)10, Co2(CO)8, Fe(η5-C5H5)2, Co(η3-C3H5), and Co(η3-C3H5)(CO)3. It can also provide qualitative explanation of the shielding or deshielding effects revealed by NMR data as well as susceptibility of an atomic region towards an electrophilic or nucleophilic attack. The topographical features of MESP and MESP-based atomic basins can be used as an auxiliary tool for the portrayal of asymmetry in molecular charge distribution.
|
|
