Barry Honig's research involves the use of computational and biophysical approaches to study the structure and function of biological macromolecules.
The guiding hypothesis of much of our research is that sequence and structural information combined with biophysical analysis can reveal the fundamental physical principles that underlie a wide range of biological phenomena. Our work includes theoretical research, biophysical measurements, the development of software tools, and specific applications to problems of biological importance. In the past few years, we have focused on two areas: 1) The incorporation of three-dimensional structural information into systems biology with particular focus on the genome scale prediction of protein-protein interactions; 2) The molecular basis of cell-cell recognition and most recently, neuronal recognition, mediated by adhesion receptors;
Integrating Structural and Systems Biology
Our long range goal is to use computational methods to predict protein-protein, protein-DNA and protein membrane interactions so as to facilitate the development and understanding of cellular networks and, ultimately, to use this information to better understand and treat disease.
Honig Abstract Slide Show
Figure 1: Overview of PrePPI. Each box represents a type of interaction evidence used. Evidence in green use structural information and evidence in blue does not. Contained in each box is a visual outline of how that type of evidence is determined. An interaction between two given proteins A and B is inferred based on some similarity or correlation (yellow double arrows) to other proteins known to interact. New features indicated by asterisks.
Figure 2: A model of a cis dimer of a full length Pcdh ectodomain. The structure is composed of two arms emanating from a membrane surface proximal to the EC6 domain. The cis dimer is the fundamental recognition unit which interacts in trans with dimers on the apposing cell surface. B The trans interactions leads to the assembly of molecular zipper.
Figure 3: Chain termination model for neuronal recognition based on zipper-assembly. Upper panel: Left – In apposed neurites shown with four identical isoforms, chain assembly is limited only by copy number. Right – A single isoform mismatch quickly terminates the chain. Lower panel: Monte-Carlo calculations demonstrate a striking dependence of Pcdh assembly size on the number of mismatched isoforms between interacting cells. Calculations assume 1000 copies per isoform. Note the step-function shape of the curve.
The genome-wide identification of pairs of interacting proteins is an important step in the elucidation of cell regulatory mechanisms. Much of our current knowledge derives from high-throughput experimental techniques as well as from manual curation of experiments on individual systems. Three-dimensional structural information has had only limited impact on this problem, in part because the number of protein sequences is vastly greater than the number of available structures. We have developed a new method, PrePPI, that uses three-dimensional structural information to predict protein-protein interactions with an accuracy and coverage that compare favorably to high-throughput experiments. PrePPI uses structural information on an unprecedented scale, an accomplishment that has been made possible through the extensive use of homology modeling combined with the exploitation of remote structural relationships. A key element in PrePPI design is a novel method that can produce structural representations (interaction models) of literally billions of putative protein-protein complexes that can then be scored in an ultrafast way. We believe that PrePPI is an important step in the widespread use of computational and structural tools to identify previously undetected protein-protein interactions and in the integration of structural and systems biology.
The current version of the PrePPI database contains about 1.35 million interactions for the human genome and is the largest database of its kind. Not surprisingly, proteins that are predicted to interact are found to have a functional relationship and we are currently exploiting this observation in a number of ways. For example, we are using PrePPI to predict novel disease-related signaling pathways, to construct disease-specific interactomes and to annotate single nucleotide polymorphisms based on the pathways they disrupt. These are all novel applications and, in each case rely heavily on structural information. As a result, they provide a unique perspective and direct connections to possible therapeutic interventions involving small molecule inhibitors.
How neurons are barcoded
Our long range goal is to understand how the structures and binding properties of cell-cell adhesion proteins determine cell-cell recognition and, ultimately, the assembly of organelles with distinct shapes. Our research program on cell adhesion is carried out in close collaboration with the lab of Lawrence Shapiro (Columbia University). Our work involves protein crystallography, biophysical measurements, in vitro cell assays, and in vivo studies, all combined with theoretical and computational simulations at multiple levels of granularity.
We have recently focused on the family of clustered protocadherins (Pcdhs) which are responsible for the phenomenon whereby neurites from the same neuron don’t cross and appear to repel one another while neurites from different neurons do not repel. The stochastic expression of multiple protocadherins on different neurons appears to provide a barcode enabling neurons to distinguish “self” from “non-self”. Pcdhs are cell adhesion proteins that initiate repulsion through an initial adhesive recognition step that then activates processes that result in neurites from the same cell moving away from one another. Figure 2A shows the structure of a two-armed Pcdh cis dimer where each arm can be different and can dimerize in trans with an identical protomer on the apposing cell surface. B. Figure 2B shows that the dimer of dimers produces a molecular zipper in the contact region between two cells. Based on the existence of this zipper we have proposed a chain termination model for neuronal barcoding (Figure 3). The basic notion is that interacting neurites from the same neuron will form large assemblies while neurites from different neurons will form small assemblies.
Portions of this work are also supported by the National Institutes of Health and the National Science Foundation.