I am a research fellow in Eric Fischer's group at Dana-Farber Cancer Institute and Harvard Medical School. The overarching goal of my research is to predict and control biomolecular assembly with a view to develop novel therapeutics and molecular biology tools.
As a graduate student in Jeffrey Gray’s group at Johns Hopkins Univeristy, I created computational tools in the Rosetta Molecular Modeling Suite to predict the structures of protein complexes. I am now using these docking and design tools to create artificial dimerization systems that can be triggered with compound treatment. Through targeted binding to E3 ubiquitin ligases, I am creating degradation systems that can be used to specifically and reversibly modulate protein levels in cells.
My hometown is New Delhi, India.
Ph.D. in Chemical & Biomolecular Engineering, 2018
Johns Hopkins University, MD, USA
B.Tech. in Biological Sciences & Bioengineering, 2012
Indian Institute of Technology Kanpur, UP, India
My current research focuses on repurposing the waste disposal system of the cell, the ubiquitin–proteasome system to selectively degrade target proteins. The three aspects of the area that I am focusing on are:
As a graduate student in Jeffrey Gray’s group, I addressed one of the principal challenges in computational protein docking: predicting binding-induced conformational changes.
Protein flexibility can be incorporated into modeling by simulating conformational selection from a pre-generated set of backbone conformations. I devised a new algorithm called Adaptive Conformer Sampling (ACS) to efficiently sample hundreds of protein backbone conformations, which when combined with a new scoring scheme developed in the lab called Motif Dock Score (MDS), allowed us to accurately discriminate native-like binding states from others. Combining ACS and MDS in RosettaDock, we were able to obtain—for the first time—successful docking of nearly 50% of flexible protein complexes with backbone conformational change of 1.5–2.2 Å RMSD. Read article.
While the above protocol deals with hetero-dimers, a majority of proteins are thought to exist as homo- oligomers. Rosetta’s symmetric docking protocol performed poorly on higher order complexes. While investigating the source of this error, I found that the binding energy landscapes of homomeric proteins differed greatly from those of heterodimeric proteins. I used the insight to simulate induced fit in tightly-packed interfaces. Combined with the discriminatory ability of MDS, the new protocol, Rosetta SymDock2 outperformed all existing algorithms by accurately predicting 61% of cyclic complexes and 42% of dihedral complexes in a diverse benchmark of various point symmetries. Read article.
To test the real-world performance of our algorithms, from 2015 to 2018, I led the lab in a community-wide blind prediction challenge called Critical Assessment of PRedicted Interactions (CAPRI) rounds 37–45. We were able to successfully model 14 out of 23 heteromer, homomer, and oligosaccharide–protein complexes with previously undisclosed experimental structures. Ex post facto analysis revealed that the new algorithms that we had developed during this cycle would have further improved our performance. Read article.
The versatility of the docking methods allowed us to predict energetically plausible models of the CoREST complex with dual inhibitor of HDAC1/2 and LSD1 developed by Philip Cole’s lab. Read article.
Key: * these authors contributed equally; † co-corresponding authors