Office: James H. Clark Center, 318 Campus Drive, Room S255, Stanford CA 94305 |
Phone: (650) 725-0754 |
Fax: (650) 723-8464 |
Email: gaurav.chopra[at]stanford[dot]edu
Education
Postdoctoral Research Fellow, Structural Biology, Stanford University - School of Medicine (2010-present)
Postdoctoral Research Fellow, Microbiology, University of Washington - School of Medicine (2010-present)
Ph.D. Institute for Computational and Mathematical Engineering, Stanford University - School of Engineering (2010)
M.S. Scientific Computing/Computational Mathematics, Stanford University - School of Engineering (2005)
M.S. Mechanical and Aerospace Engineering, University of California, Irvine - The Henry Samueli School of Engineering (2003)
B.Tech. Mechanical Engineering, Indian Institute of Technology (IIT), Delhi, India (2002)
KobaMIN: Knowledge based minimization server for protein structure refinement.
Research projects
A SHOTGUN APPROACH TO DISCOVER INFECTIOUS DISEASE THERAPEUTICS
Structure-based drug discovery is typically limited to, and by, screening a compound library against one protein structure associated with a particular disease (target) to identify inhibitor leads that eventually will lead to drugs that treat the disease. We will develop a comprehensive high throughput computational drug discovery platform by enhancing a novel technique for a dynamic, fragment based screening of a complete set of FDA approved synthetic and natural small molecule compounds against the three dimensional structures of multiple protein targets from infectious disease causing pathogens, followed by prospective in vitro and in vivo experimental verification done by our collaborators. Our approach is to computationally curate an intelligently compiled compound library, select for inhibitory function in a set of target disease proteins, and avoid side effects. To maximize the chance of efficacy and safety in humans, the selected small molecules should be existing drugs or natural compounds. The probability of finding pharmacologically active compounds increases greatly by targeting multiple proteins related to a disease or in different disease associated pathogens. We will use this entire set through our high throughput fragment-based docking with dynamics approach to find more alternative drugs to solve the problem of drug resistance, especially during a pandemic. We also believe that such a multi-targeted hybrid computational and experimental approach is necessary to find drugs efficiently for novel synthetic pathogens. Our predictions for discovery of broad-spectrum herpes virus inhibitor have led to high hit rates that are unmatched by high throughput screening or traditional wet experiments (paper in preparation).
We will initially focus on benchmarking our method with the known drugs and targets. Using this available information along with additional drug-target binding experiments we will develop confidence measures to assess our predictions and we will evaluate the efficacy of new drug-target interactions in order develop novel therapies. Instead of modeling the low resolution protein structures, we will screen the entire compound library with the high resolution protein structures of pathogens such as Burkholderia pseudomallei, Burkholderia mallai, Salmonella etc. We will make sure that our computational predictions rank the known drugs to the corresponding binding sites with high scores. Failing this, we will train our scoring function based on Kd affinities for the known drugs using machine learning techniques; this will ensure that we pick the relevant drugs from the dataset of known drugs and binding sites for all pathogens. We will also do Kd studies on the top 50 high scoring drugs which are not known for this benchmark set; this will create a "gold standard" set of new and previously known compounds. These results will be used to iteratively calibrate our computational methods to make increasingly accurate predictions.
PROTEIN STRUCTURE REFINEMENT
We modeled the effect of solvent (water) environment on protein structure refinement using physics-based potentials with implicit and explicit water models as well as the knowledge-based (KB) statistical potential functions derived from high-resolution X-ray diffraction data of protein structures. Using energy minimization and molecular dynamics, we tested the physics-based and KB potentials on a large set of protein structures. The physics-based implicit solvent models demonstrated large magnitude of refinement but the KB potential was more consistent. We have prospectively verified the accuracy of the KB potential in the refinement category at the eight and ninth world-wide experiments on Critical Assessment of techniques for protein Structure Prediction (CASP). Evaluation of our protocol on all models generated at previous CASPs indicates that this simple, consistent and computationally efficient refinement protocol is a natural "end" step for protein structure prediction.
POLARIZATION AFFECTS BIOLOGICAL PROCESSES
Since the pioneering work on the consistent force field developed by Shneior Lifson, more than 50 years ago, the model of an atom has been a nucleus with a partial charge. We believe the time has come to move to a more realistic representation of an atom as a nuclei and an exponentially distributed zero mass electron cloud around it. This has been implemented in a recently introduced state-of-the-art quantum general purpose quantum mechanical polarizable force field (QMPFF3) fitted solely to high-level quantum mechanical data at MP2/cc-pVTZ level with a simple model correction using CCSD(T) data, allowing the effect of polarization to be correctly modeled. Accurate description of the water structure around the solute of interest could improve our understanding of various biological processes such as protein folding. We study the structure of polarized water around polarized hydrophobic solutes of varying sizes (methane, benzene, cyclohexane and Buckminsterfullerene) to show that polarization induces a strong hydrophobic effect; this has been under-represented by the limitations due to approximate modeling of atomic interactions in the empirical force fields widely used for the past decades. Our sensitive method to detect surface roughness shows that the hydrophobic effect is much stronger at short- and long-range for QMPFF3 compared to classical force fields simulations for polarized Buckminsterfullerene. The major conclusion from this study is that a quantum mechanical force field increases the strength of the hydrophobic effect; this could have a profound affect on protein folding.
Our results with nonpolar solutes give us enough confidence that polarization is an important component which should not be neglected in empirical force fields and should be modeled correctly for biological systems. We will use QMPFF3 to study the affect of polarization on biological systems like proteins and protein-ligand complexes in non-homogenous solvent simulations (for enhanced sampling). To this end, we are making several advances to increase the accuracy and speed of this polarizable force field.
