The AbDesign computational pipeline for modular backbone assembly and design of binders and enzymes

Abstract

The functional sites of many protein families are dominated by diverse backbone regions that lack secondary structure (loops) but fold stably into their functionally competent state. Nevertheless, the design of structured loop regions from scratch, especially in functional sites, has met with great difficulty. We therefore developed an approach, called AbDesign, to exploit the natural modularity of many protein families and computationally assemble a large number of new backbones by combining naturally occurring modular fragments. This strategy yielded large, atomically accurate, and highly efficient proteins, including antibodies and enzymes exhibiting dozens of mutations from any natural protein. The combinatorial backbone‐conformation space that can be accessed by AbDesign even for a modestly sized family of homologs may exceed the diversity in the entire PDB, providing the sub‐Ångstrom level of control over the positioning of active‐site groups that is necessary for obtaining highly active proteins. This manuscript describes how to implement the pipeline using code that is freely available at https://github.com/Fleishman‐Lab/AbDesign_for_enzymes.

Assessing and enhancing foldability in designed proteins

“…The current study is motivated by our ongoing efforts to develop a reliable strategy for de novo enzyme design. This strategy is based on recent methods developed in our lab to design new backbones through the modular assembly of large (60-150 amino acid) backbone fragments of natural enzymes [25,29,30]. Modular assembly and design was successful in assembling elements from existing protein families, generating accurate antibodies [27] and ultrahigh specificity binders [24] and enzymes [25] with as many as 100 mutations from any natural protein.…”

Recent advances in de novo protein design: Principles, methods, and applications

“…A recent method called AbDesign ( 66 ) seeks to mimic natural homologous recombination. In contrast to other methods, Abdesign uses larger segments and relies on the similarity between members of the same protein family to facilitate backbone sampling.…”

High-throughput screening, next generation sequencing and machine learning: advanced methods in enzyme engineering

“…For this reason enzyme engineers have started to regularly use those methods to improve the outcome of directed evolution and enzyme engineering campaigns. 87–97 Machine learning can improve the efficiency of downstream experimental studies thereby adding value and complimenting purely experimental bioengineering approaches. Moreover, the exponential increase in DNA sequencing throughput presents a significant opportunity to combine state-of-the-art computation and machine learning with large biological datasets in an attempt to learn sequence-function maps in protein sequence space.…”