Research Interests

Structural Modeling of MHC-bound Antigen Peptides

In the context of cancer immunotherapy it is key to be able to identify which peptides specific to the tumor may bind to the MHC to be presented to the immune system (neoantigens). Even more important is to predict which of these neoantigens will trigger a significant T-cell response.

While sequence-based predictors show some success for MHC binding, predicting antigenicity remains challenging without structural information. In collaboration with V. Zoete at SIB and CHUV, we use a combination of docking, protein modeling and molecular dynamics simulations to model the structure and dynamics of MHC-bound peptides. A reliable structural model will be the stepping stone to identify determinants of TCR binding, and thus of antigenicity.

In Situ Data Analytics for Next Generation Molecular Dynamics Workflows

Next-generation high-performance computing (HPC) systems will have dramatically larger compute performance than current systems do, which translastes into the ability to create huge amounts of MD data. But since storage systems will not increase their bandwidth concommittently, data storage and analysis will become the main bottleneck. The Analytics4MD project tackles the data challenges of MD simulations at the exascale through (1) creation of novel data analytics algorithms ideal for in situ data analysis of relevant structural molecular properties, (2) definition of MD-based machine learning (ML) techniques to automatically identify the molecular domains where the properties reside at runtime, and (3) integration of both algorithms and techniques into MD workflows at the extreme scale.

Analytics4MD is a collaborative effort between the University of Tennessee Knoxville, the University of Southern California, the University of New Mexico, and Weill Cornell Medicine, funded by an combined grant of the American National Science Foundation.

The T Cell Receptor as a Mechanosensor

T lymphocyte response is determined by the interaction of the T cell receptor (TCR) with peptide (p) antigens presented by MHC molecules. While he molecular determinants of recognition at the TCR-pMHC interface have been extensively studied, the mechanisms by which the activation signal is transmitted into the T cell are still to be discovered. Understanding and optimizing TCR signaling is key to the development of therapeutic approaches such as adoptive transfer of genetically engineered T cells for cancer immunotherapy. It has recently emerged that the TCR functions as a mechanosensor that recognizes agonist antigens when subjected to cytoskeletal forces. However,  the molecular mechanisms underpinning the mechanosensor function remain unknown. We use steered molecular dynamics to study the effect of external forces on the TCR bound to agonist or non-agnonist antigens. As nonequilibrium simulations require many replica, this project involves large-scale computational resources, such as the Piz Daint supercomputer at the Swiss National Supercomputing Centre.

Quantifying Allostery in Proteins: the Thermodynamic Coupling Function Analysis Method

Going beyond the usual two-state models of allostery, in collaboration with M. LeVine and H- Weinstein at WCM, we introduced a statistical mechanical formalism to describe rigorously the coupling between two collective variables that represent key conformational changes in a protein, such as ligand binding and activation. We show that allosteric coupling is best represented as a two-dimensional thermodynamic coupling function (TCF).

We combined the TCF formalism and a Markov state model analysis of the human dopamine transporter (hDAT), revealing a non-trivial thermodynamic coupling landscape between the sodium release and intracellular gating steps.

We make available online AlloDeco, a fast version of the TCF analysis to decompose allosteric coupling based on coarse-grained Gaussian Network models.
allostery landscape

Neurotransmitter Sodium-coupled Transporter Proteins

                            in membrane The glutamate transporter GltPh is a homolog of mammalian excitatory amino acid transporters (EAATs) that mediate glutamate re-uptake after discharge at the neuronal synaptic cleft. In the transport cycle, the three homotrimeric transport domains (blue) undergo an elevator-like motion relative to a scaffold domain (wheat). With a team at Weill Cornell, we reported in Nature significantly higher transport rates for a GltPh construct in which key residues were mutated to mimic human EAAT1. The mutant adopted a novel “unlocked” conformation (PDB 4X2S) that for which MD simulations showed to be stable only if hydrophobic molecules such as lipid tails insert between domains.

We are currently expanding this work to quantify the contributions of single residues to the stability of several intermediate states. In particular, we are interested in understancing the allosteric coupling between the closure of the ligand binding site by the HP2 loop (the "elevator door") and the inward elevator motion. This coupling that prevents transport of sodium without substrate ("sodium leak"), is essential to the symport function. We are also investigating the role of protonation of key acidic residues in the ligand binding site.

Past Research

Please go to the next page