The development of computational methods for designing antibodies and other proteins.

摘要

Proteins are polymers of amino acids that have essential and diverse roles in organisms, including: structure (e.g. actin), catalysis of chemical reactions (e.g. cytochrome p450), and signaling (e.g. insulin-like growth factor I). Given the wide range of functions that proteins fulfill in nature, there is much interest in utilizing them for human needs in many areas, such as biofuels production, materials, and medicine. However, nature rarely provides a protein that is perfect for a specific human application, necessitating the use of engineering methods to improve or create desired properties. Computations are an essential tool for the de novo design of proteins. This dissertation focuses on the use of antibodies as a model protein system to develop de novo protein design methods. Due to their many useful experimental and medicinal applications, antibody structures and their natural mechanisms of generation have been extensively studied. They are an excellent system for learning de novo protein design principles, as their structures have many modular features and their functions are limited to binding, not catalysis. Antigen binding by antibodies is primarily driven by the complementarity determining regions (CDRs). Models of the possible backbone conformations of the CDRs (i.e. canonical structures) were generated. These models were used in the Optimal Complementarity Determining Regions (OptCDR) method to allow the de novo design of antibody CDRs to bind any specified antigen epitope. Next, a database of Modular Antibody Parts (MAPs) analogous to the human germline genes used to make antibodies was created and shown to be able to predict antibody structures with a high degree of accuracy. Analysis of calculations involving OptCDR, MAPs, and other work outside the scope of this dissertation suggested that the computational protein engineering methods currently in use needed to be improved. This led to the development of the Iterative Protein Redesign & Optimization Suite of Programs for the (re)design of proteins. Numerous collaborations have been established to experimentally validate the computational predictions and the research is progressing towards the de novo design of fully human antibodies.