Modeling protein interactions to interpret genetic variation

A grand challenge for biology is to maximize the fundamental insights from high-throughput sequencing, which has become rapid and inexpensive. Structural information on proteins and their interactions is essential for understanding the effects of genetic variation. A vast amount of information on single amino acid variants (SAV) will be available from eukaryotic and prokaryotic organisms. We will develop an integrated approach for large-scale prediction of protein structures and their association. A database of predicted structures and complexes for model organisms will be established upon which genetic variants will be mapped and their phenotypic effect assessed.

The Objectives of the proposed research are: (1) to develop high-throughput structure-based methods to predict interactions of experimentally determined and modeled proteins; (2) to develop advanced methodology for high-throughput modeling of individual proteins; (3) to generate genome-wide database of protein structures and protein-protein complexes for model organisms; and (4) to assess phenotypic effects of genetic variation. Approaches will be developed to discriminate non-interacting from interacting proteins, and to model the structures of protein-protein complexes, based on similarity to experimentally determined protein-protein complexes and on properties of the intermolecular energy landscape. A novel approach for fold detection will extend the number of proteins that can be modeled. A pipeline will be developed to integrate protein structure prediction with the prediction of protein-protein complexes, and use structure-based approaches to predict the effects of SAVs. This collaborative proposal combines highly complementary areas of expertise of the US team, on high-throughput modeling of protein-protein interactions, and the UK team, on protein structure prediction and SAV effects.

his project is a collaboration between Prof Vakser's group in Kansas (USA) and Prof Sternberg's at Imperial College with the following research goals.

*DEVELOP HIGH-THROUGHPUT STRUCTURE-BASED METHODS TO PREDICT INTERACTIONS OF EXPERIMENTALLY DETERMINED AND MODELED PROTEINS. We will develop approaches for modeling the structures of protein-protein complexes, based on the similarity to experimentally determined protein-protein complexes (templates). Comprehensive benchmark sets of interacting and non-interacting modeled proteins will be generated based on the datasets of experimentally determined protein complexes.

*DEVELOP ADVANCED METHODOLOGY FOR HIGH-THROUGHPUT MODELING OF INDIVIDUAL PROTEINS. We will model the structures of predicted interacting proteins. The Phyre2 structure prediction server (> 70,000 users worldwide) implements the state-of-the-art approach to protein modeling using sequence-based fold detection of remote homology. Recently a major breakthrough occurred in prediction of residue contacts from sequence. We will develop this novel approach extending the number of proteins that can be predicted.

*GENERATE GENOME-WIDE DATABASE OF PROTEIN STRUCTURES AND PROTEIN-PROTEIN COMPLEXES FOR MODEL ORGANISMS. We will develop a pipeline to integrate protein structure prediction with the prediction of protein-protein complexes and link servers in the two groups for use by the community. We will generate a database of structurally refined protein complexes for model eukaryotic and prokaryotic organisms. The database will be made available to the research community.

*ASSESS PHENOTYPIC EFFECTS OF GENETIC VARIATION. A pipeline will be developed for users to map amino acid variants onto the structures and complexes and use structure based approaches to predict phenotypic effects. The predicted effect on structure will involve modeling acceptable conformations for the variant side-chain and evaluating the change in interactions.

A grand challenge for biology is to maximise the fundamental insights from high-throughput sequencing which has become rapid and inexpensive. Structural information on proteins and their interactions is essential for obtaining insights on genetic variation. A vast amount of information on single amino acid variants (SAV) will be available from eukaryotic and prokaryotic organisms. We will develop an integrated approach for large-scale prediction of protein structures and their association. This project is a collaboration between Prof Vakser's group in Kansas (USA) and Prof Sternberg's at Imperial College. A database of predicted structures (from Phyre2 via Imperial) and complexes for model organisms (from GWIDD via Kansas) will be established upon which genetic variants will be mapped and their phenotypic effect assessed (from SuSPect2 via Imperial)). We will now identify those groups that will benefit from this research and in what way they will benefit.

ACADEMIC - Many of the users of the resources developed under this grant will be academic groups and the results of their use of these resources will advance their research leading to economic and social benefit. The current users of the present resources at Imperial and at Kansa are international and spans diverse researchers in bioscience who require information about protein structure, interactions, function and the effect of genetic variation. Feedback from users has shown that these predictions can have a transformative effect on their research moving their conceptualisation into detailed consideration of the molecule at the three-dimensional atomic level. There are numerous application areas. One major application area is the identification of novel targets for pharmaceutical intervention. Structure guides both the design of small molecules and bio-therapeutics, such as monoclonal antibodies. The consequence of the design of novel pharmaceuticals has clear health and commercial benefit. A second application area is the agricultural sector. Similar considerations apply to animal health as for the pharmaceutical industry. In addition, genome information can be helpful in selective breeding of crops and fruit. The biotechnology and bio-energy sectors can focus on the modification of biological pathways and information about the structure and function of genes can inform these studies.

PUBLIC SECTIOR - Agencies involved in public health and food security are expected to use the resources. For example the location of a mutation on the surface of a human, animal or plant pathogen could be mapped to provide insight into structure/function relationships. This will impact on health and well-being.

POLICY MAKERS AND THE PUBLIC - Via open days at Imperial College, members of the general public will see demonstrations of protein modelling. This will highlight an area of research - bioinformatics - of which they may not have been aware. In particular the Imperial Festival is an annual event which attracted over 15,000 in 2016. From the policy side, Imperial invites to the Festival representatives from professional membership bodies, local and central government, higher education bodies including other university senior staff, and research funders. We will continue to give invited lectures to groups other than researchers. Previously, Prof Sternberg has addressed the Prince's Trust and a meeting at Brighton linking Art and Science.

SCHOOLS - In talks to schools, the Phyre2 server is described as a web-based resource for use by the community. This always has a major impact on the audience. Students are impressed by the Phyre2 usage figure - over 1.5 million hits. The opportunity to develop a resource used by so many other scientists excites students as a highly worthwhile activity.