Research at the Furman Lab

Protein-protein interactions are at the heart of any functioning cell and organism. We are interested in elucidation of the structural details that allow for the complexity observed in protein interactions and function.

We use computational tools, including structure-based computational prediction and manipulation of specific interactions, analysis of evolutionary signals hidden in sequences, and large-scale integration of this data by machine-learning approaches. In particular, we exploit structures and their surrounding energy landscape sampled by our simulations to learn about the binding process of protein interactions.

One of our main current research interests is directed towards the accurate modeling of peptide-protein interactions, since these play an increasingly important role in mediating a significant fraction of protein-protein interactions. In addition, they allow us to tackle the modeling of extensive backbone flexibility in a defined setup.

High-resolution modeling of protein-protein interactions

We use RosettaDock to create high-resolution models of protein-protein interactions. RosettaDock is a well-established docking protocol for the modeling of protein complexes starting from the free monomers that explicit models conformational changes at the interface that occur upon binding. With this protocol we have been able to create high-resolution models in many of the CAPRI rounds. Thus, RosettaDock can be used to create highly accurate models of protein complex structures that serve as starting point for the targeted manipulation of interactions by structure-based computational design, thereby opening an exciting new field of applications. In our current work, we extend our existing protocol to a variety of related applications, such as the prediction of protein-peptide binding, protein motion and binding specificity.

Current research focus:

Our current research concentrates on:
(1) Characterization, modeling and manipulation of peptide-mediated interactions: in search for principles that govern peptide-mediated protein interactions, we have created a cruated set of peptide-protein complex structures (PeptiDB), and compared features of these interactions to globular protein-protein interactions.

(2) Insights from this study laid the ground for the development of Rosetta FlexPepDock, a protocol for the accurate modeling of peptide-protein complex structures. Go to our new FlexPepDock server

(3) Application of structure-based characterization and manipulation to specific protein interactions of biomedical and biotechnological interest: In particular, we have performed a detailed analysis of the cohesin-dockerin interaction to study determinants at the interface that contribute to binding affinity, pormiscuity and specificity. This is a collaboration with the Bayer group at the Weizmann institute. .

(4) Modeling of protein motion by implementing computational geometry approaches into Rosetta: Application of tools from the robotics field could well help us model protein motion more efficiently. Based on this reasoning, we have developed Rosetta PathRover in which we implemented the RRT (Rapidly growing Random Trees) algorithm into the Rosetta framework to efficiently model protein motion. Using this approach, we have characterized protein motion for a range of different systems.

(5) Generalization of RosettaDock to model backbone conformational changes upon binding: We have evaluated whether structures solved for homolog proteins could be helpful for improved docking with backbone flexibility. This study found that improved predictions are mainly obtained when the structure of the homolog is missing a flexible part, and therefore prevents clashes.

(6) Analysis of energy landscapes sampled in RosettaDock and other simulations: We have compared near-native funnels with funnels in other regions of the energy landscape sampled by RosettaDock. We found characteristic features and have developed FunHunt, a classifier that identifies near-native minima. Go to our FunHunt server

(7) Conserved protein domain repeats and their functional role: Our initial studies of Cohesin-dockerin interactions lead to an interesting finding: Cohesin repeats on the scaffoldin are exceptionally well conserved in sequence. What is the functional implication of this finding? Will proteins now aggregate? We have performed an in depth study of protein repeat conservation and found significant differences between eukaryotes and eukaryotes, and identified the state of two repeats as the critical step. We are now conducting a large scale analysis of protein repeats and their function, struture, and history.

Our Goal

We have established, and are further developing, state-of the art, leading tools for the structure-based computational manipulation of proteins and their interactions. In the long term, we will bring those techniques to a point where we can (1) reliably apply them to large scale data and address general questions at the level of system biology; (2) consolidate routine applications for a range of interesting specific biomedical and biotechnological systems. Towards this goal, we have established, and continue to establish fruitful collaborations with experimental groups, which allow the cycling between modeling and experiment that is critical for our advancement in both our understanding of the basic underlying principles of protein interactions, as well in their applications.