In silico design of antibodies for biomedical applications

Abstract

Proteins are large macromolecules constituted by amino acids that are responsible for most of the biological processes within a cell. Proteins showing high complementary affinity may bind forming protein-protein complexes. In this context, antibodies are proteins that recognize abnormal particles in the body (known as epitopes), and are elicited by means of random recombinatory events followed by strict screening selection processes. Along their production, antibodies can be modified by mutation events leading to potent antibody variants.

In this sense, there is an industrial and biomedical interest for the artificial optimization of antibodies. The rise of the computational era together with the deeper understanding of structural biology allowed the design and implementation of predictive algorithms for simulating the effects of mutations in protein-protein complexes. This process usually involves, among others, the prediction of changes in Gibbs free energy upon mutation and the use of other computational simulations for unveiling motions and binding patterns, such as Molecular Dynamics and Monte Carlo techniques.

During this thesis, we have developed and implemented predictive algorithms focused on the design of potent antibody variants. We developed UEP, an open-source code for predicting the effects of mutations in protein-protein complexes. UEP differs from the state-of-the-art and employs other sources of knowledge rather than experimental binding affinity determinations upon mutation. Moreover, we designed a PELE protocol to simulate the binding affinity of antibodies against hypermutated HIV-1 viral isolates. Finally, we describe three different computational workflows for antibody optimization. We particularly focused on the challenge of increasing the binding potency of the N6 antibody, one of the best antibodies against HIV-1. Each computational workflow has been evaluated experimentally by our collaborators from Irsicaixa, and such combined computational and experimental effort resulted in the design of an improved variant of the N6 antibody against HIV-1.