Structure-based computational design of antibody mimetics: challenges and perspectives

Abstract

The design of antibody mimetics holds great promise for revolutionizing therapeutic interventions by offering alternatives to conventional antibody therapies. Structure-based computational approaches have emerged as indispensable tools in the rational design of those molecules, enabling the precise manipulation of their structural and functional properties. This review covers the main classes of designed antigen-binding motifs, as well as alternative strategies to develop tailored ones. We discuss the intricacies of different computational protein-protein interaction design strategies, showcased by selected successful cases in the literature. Subsequently, we explore the latest advancements in the computational techniques including the integration of machine and deep learning methodologies into the design framework, which has led to an augmented design pipeline. Finally, we verse onto the current challenges that stand in the way between high-throughput computer design of antibody mimetics and experimental realization, offering a forward-looking perspective into the field and the promises it holds to biotechnology.

Keywords: de novo design; deep learning; machine learning; protein engineering; protein structure.

Fig. 1

Protein structure scaffolds of the main antibody mimetics. Structures are shown in cartoon model, where the framework and binding domains are represented in gray and orange, respectively. The accession codes for each structure in the PDB and the amino acids on the binding domain are the following: 8DA4 (Affibody), residues 9–11, 13, 14, 17, 18, 24, 25, 27, 28, 31, and 32; 4N6T (Affimer), residues 60–71, and 98–100; 5AEI (dArmRP), residues 65–83, 108–126, 149–168, 192–210, 233–252, 91–101, 115–121, and 141–156; 1N0S (Anticalin), residues 31–43, 62–65, 90–93, and 117–123; 2XEE (DARPin), residues 43, 45, 46, 48, 56, 57; 7S5B (Mini‐protein), residues 1–16; 1TEN (Monobody), residues 813–818, 827–831, 840–846, 862–867, and 877–882; 1I3V (Nanobody), residues 26–32, 52–62, and 105–116.

Fig. 2

Computational workflow strategies for the design of antibody mimetics. (A) Initial step consists of obtaining the structural and physicochemical characteristics of the antigen target interface (or antigen–antibody interface), and if possible, to identify key‐residues for recognition. This information will be selectively passed on the design strategy used. (B) The Motif Grafting method requires structural data of antigen–antibody interface. Selected regions of the paratope that are considered important to recognition will be attempted to be grafted onto a variety of scaffolds. A scaffold that can withstand grafting will have its interface and topology sequence optimized to confer the desired affinity and protein stability, respectively. This example shows the grafting of CDR from conventional antibodies onto a VHH. (C) The Docking & Design method requires structural information from the antigen only. Scaffolds from a pre‐triaged structural library will be attempted to dock onto the antigen target surface, considering their shape complementarity. Once the two proteins are docked, mimetic will undergo interface optimization to achieve high affinity, and subsequently optimization of its scaffold (where the pre‐optimized interface sequence is kept fixed) to confer protein stability. It is exemplified here by attempting to dock the affimer, anticalin and DARPin mimetic classes. (D) De Novo Design creates a topology from scratch. It uses only structural data from the antigen only. However, a higher chance of success is achieved when antigen–antibody structural data is available. In the latter, a structural motif carrying the key‐interacting residues of the antibody is used as a starting point. Proteins are designed to vehicle the chosen motif, as close as possible from its native form. The example shows the building of a miniprotein. (E) As in the De Novo Design, one can use either structural data from the antigen only or antigen–antibody, if available. The example depicts the generation of an affibody using ANN, that uses as starting point selected key‐interacting residues from the antibody. Finally, it is worth noting that, even when key‐interacting residues from the antibody are used as starting point, they can be optimized (mutated) to generate a molecule with enhanced binding affinity (all presented examples are merely illustrative and therefore do not belong to any published study).