Research


Cells are miniature machines, as are proteins and other molecules of life. The function of those machines, and the design of medicines to fix them when they break, depends on their three-dimensional structure. We use cutting-edge computational approaches, together with experimental data, to understand this structure, how it changes over time, and how we can exploit it for drug design.

We aim not only to develop new methods but also to have a real impact on biology. Some members of the lab thus delve deeply into the study of specific biological systems. We are particularly (though certainly not exclusively) interested in G protein-coupled receptors (GPCRs). GPCRs are membrane proteins that represent the largest class of drug targets: about a third of all drugs act by binding to these receptors and causing or preventing changes to their structure. Recent experimental breakthroughs—recognized by the 2012 Nobel Prize to Brian Kobilka, our Stanford colleague and collaborator—have revealed the structures of many GPCRs. We use computational methods to reveal the dynamics and mechanisms of these proteins, in collaboration with leading academic experimentalists and drug companies.

Most of the projects currently underway in the lab involve one or more of the following approaches, but the techniques we employ are continually evolving, and we always welcome new ideas.

  • Molecular simulation: Molecular dynamics simulations predict the atomic-level motions of proteins using basic physical principles. They have become much more powerful in recent years, thanks to advances in computer power, algorithms, and chemical models. They can now be used to reveal the workings of processes such as drugs binding to their targets, protein folding, or the structural changes that underlie protein function.

  • Cellular modeling using fluorescence microscopy data: Cells are miniature machines, and biology textbooks are full of cartoons showing how these machines work, but how do the molecules within a cell actually move around to execute the cell’s “programs”? How do these processes go wrong in diseased states, and how can we fix them? To address these questions, we are building spatially resolved cellular simulations informed by real-world fluorescence microscopy data. This requires both new methods for analyzing microscopy imagery and new cellular- scale simulation techniques.

  • Predicting protein complexes: Proteins function largely by associating with one another to form complexes, but determining the structure of these complexes experimentally is often challenging. We aim to predict such structures using statistical learning techniques in concert with various types of data, including genetic sequences of similar proteins and known structures of other protein complexes.

  • Machine learning and statistics: We use statistical methods, including modern machine learning techniques, to infer structural models from experimental data, to analyze simulation results, and to synthesize data from different sources.