Rational design relies on a deep understanding of the relationship between protein structure and function to make targeted modifications to specific amino acids or protein domains with the aim of achieving a desired protein behavior. This approach can make small changes to an existing protein, like surface mutations to improve an enzyme’s solubility, or can build entirely new proteins from existing parts, such as the construction of a chimeric antigen receptor on T-cells. Rational design can be employed to achieve a wide variety of different engineering goals and its targeted nature means a more streamlined design-build-test cycle, often significantly reducing the number of designs required to achieve the desired outcome.

Directed evolution mimics natural selection in the laboratory and involves creating diverse libraries of protein variants through random or focused mutagenesis and selecting for variants that exhibit ideal properties. It is an iterative process that allows for the gradual improvement of protein characteristics over multiple rounds of design and selection. One example of this technique is the evolution of an enzyme to increase its heat tolerance and improve its utility for biomass conversion during biofuel production. Directed evolution is often the best choice when the protein of interest is an existing protein with little structural or functional information available. In this situation, the path towards achieving the desired outcome is often unclear so this approach allows for screening of a wide mutation space to identify initial design leads for refinement in later iterations.

Computational approaches have made great strides in recent years, enabling protein engineers to model protein structures and reduce dependence on experimental structural determination. AlphaFold2 and RoseTTAFold are two examples of deep learning approaches that have dramatically improved the structural modeling space. Additional deep learning approaches are under development that allow for computational design, or de novo design, of proteins. This offers the potential to create entirely new, non-natural proteins with specified structure and function based on knowledge of protein folding energetics and structure-function relationships. De novo design is most successful when building proteins with greater than 80% novelty, such as the design of a fully non-natural inhibitor of PD-L1. Refinements to existing proteins or design of multi-domain chimeras are often best accomplished by other methods.