January 2019: Funding proposal to the European Commission submitted. Please contact the Project Manager for further information on this proposal.
October 2018: Partner proposal shared with interested members.
Why is this important?
Antibody therapeutic discovery is guided by predicting and understanding the structure of the antibody alone and in complex with its antigen. Structural information is of particular importance during drug discovery to explain and improve affinity, specificity and developability liabilities of the therapeutic candidate. In addition, antibody structure information will take an increasing role during the early stages of discovery as a consequence of large-scale experiments and analyses based on sequence data, e.g. next-generation-sequencing (NGS) of antibody repertoires. The use of large datasets is rising in popularity and nearly every immunisation-based Ab campaign is now accompanied by generation of an NGS dataset. Since Ab structure determines Ag specificity, such large scale structural characterization of an entire repertoire could provide information on the specificity of individual Abs at a much earlier stage than is currently possible.
Determination of the structure and function of antibody complexes is critical for drug development, but purely experimental studies are slow, expensive, and sometimes fail. In the last few years, computational modelling of antibodies alone and docking of the antibody-antigen complexes have developed tremendously but they are still some way from being used routinely to influence therapeutic discovery due to their unreliability in providing accurate and consistent results.
The use of surrogate screening assays and computational tools for the prediction of ‘developability’ parameters such as chemical and structural stability, aggregation, PK and viscosity is a potentially efficient and effective way to help rank candidate molecules. There has been large number of publications over the last 5 – 10 years, describing assays to predict different aspects of ‘developability’, many of which require relatively small amounts of protein (micrograms) to characterise and are designed to handle many candidate antibodies at once. These include, antibody solution properties, such as, solubility, viscosity and self-association, e.g. CIC, CSI-BLI, AC-SINS; physical/chemical stability, e.g. thermal stability by DSC/DSF, charge heterogeneity by cIEF; and assays to predict atypical clearance in vivo due to poly-specificity/non-specific binding, e.g. baculovirus binding, PSR assay.
Despite this, our confidence in the ability to predict ‘developability’ risk early is limited by the low number of molecules with associated data and an agreed standard way of measuring this data across the industry. In addition, key pharmacological properties such as human and cyno PK (non-specific clearance), viscosity and solubility, and an understanding of whether these endpoint data could be predicted using biophysical surrogate data is lacking.
What will the project achieve?
The project will enhance and optimise existing computational tools for modelling and docking antibodies and antibody-like molecules, and validate and improve usability of these novel computational tools in the context of antibody drug discovery.
The project will create an industry-wide repository of biophysical and PK information for antibodies which will be used to validate and refine existing in silico predictive models of ‘developability’ risk and generate new models.
How will the project do this?
The project will develop and optimise existing computational tools (ABodyBuilder, HADDOCK, GROMACS) specifically for antibodies, antibody-like molecules and their complexes by using novel algorithms, enhanced sampling methods and incorporating experimental data to guide modelling and the selection of relevant binding conformations.
The project will compile sequence, structure, biophysical and endpoint information on antibodies from member pharmaceutical companies with a view to better predict ‘developability’ risk. In relation to this proposal, Pistoia Alliance is facilitating a closely related initiative aiming to conduct a detailed literature review to extract well curated PK data on fully characterised antibody molecules in order to build a ‘gold standard’ set on which to base further work. This initiative resulted from the EMBL-EBI Industry programme workshop on PK prediction of Biologics held in June 2018.