Data-driven computational protein design

Abstract

Computational protein design can generate proteins not found in nature that adopt desired structures and perform novel functions. Although proteins could, in theory, be designed with ab initio methods, practical success has come from using large amounts of data that describe the sequences, structures, and functions of existing proteins and their variants. We present recent creative uses of multiple-sequence alignments, protein structures, and high-throughput functional assays in computational protein design. Approaches range from enhancing structure-based design with experimental data to building regression models to training deep neural nets that generate novel sequences. Looking ahead, deep learning will be increasingly important for maximizing the value of data for protein design.

Figure 1.

Overview of data-driven design. (A) Design can be based on protein structures, sequences, or experimental properties. (B) Raw data are sometimes processed to give derived constructs such as TERMs, van der Mers, single-residue frequencies, residue-residue covariation strengths, or single-mutant fitness values. (C) Raw or processed data are used to generate different types of models, e.g. models based on statistical analysis, regression, or neural networks, which can describe first-order, second-order, or possibly higher-order contributions from protein residues. (D) Models can be used to generate one or more new protein sequences or a quantitative description of a fitness landscape that can be used for design. Data-based methods can also guide library screening (purple) or be combined with methods that involve all-atom modeling (not shown here).