Research in the Gray Lab focuses on computational protein structure prediction and design. We develop computational tools to predict the structure of antibodies, glycans, membrane proteins, antibody–antigen complexes, protein–protein complexes, and protein–mineral-surface complexes. As a member of the RosettaCommons we also co-develop the Rosetta biomolecular modeling suite.

Read more about each of our focus areas below:

Protein-Protein Docking

Protein–protein interactions are central to biology. High-resolution structures of protein complexes provide insight to the molecular mechanism of function. While structures can be determined experimentally, such methods have limited throughput and are not viable for all proteins. Thus, we develop computational tools to model protein–protein interactions, as an alternative approach.

Membrane Proteins

Proteins embedded in cell and organelle membranes constitute 30% of proteins and are targets for 60% of drugs. However, knowledge of their structures is sparse. Our goal is to develop tools to investigate the structures and interactions of proteins in the membrane toward understanding their role in human health.

Antibody Modeling

The advent of next-generation sequencing has enabled the feasible determination of numerous antibody sequences. While this information is useful, structures are necessary for understanding antibody–antigen interactions, and traditional methods cannot determine the structures of all sequenced antibodies. To bridge the sequence–structure gap, we develop a computational structure prediction method: Rosetta Antibody.

Antibody-Antigen Docking

Determining the antibody–antigen binding mode is necessary for the rational design of vaccines and the development of antibody therapeutics. Experimental methods can be used, but are limited by their throughput and expense. As with protein–protein docking, we develop antibody–antigen specific computational docking methods to provide an alternative to experimental approaches.


Carbohydrates compose the most abundant class of molecule on the planet, yet a structural understanding of their role in biomolecular pathways is limited. Carbohydrates pose unique modeling challenges in sampling, scoring, and nomenclature, relative to the modeling of peptides. We have recently developed RosettaCarbohydrate, a new framework for modeling saccharide and glycocongugate structures to overcome these challenges.


To better treat biomineralization diseases or design novel biomimetics, we must first understand the interactions of proteins with mineral surfaces. Toward this end we have developed RosettaSurface, an algorithm designed to broadly sample conformational space and identify low-energy structures. We are also working on thermodynamically-rigorous simulations to predict free energies of binding peptides to surfaces.