Evolution is often visualized as a semi-random survival-of-the-fittest process, with a population stochastically but gradually climbing toward the top of a fitness landscape. Although individual proteins can be computationally designed or experimentally evolved to exhibit a desired structure or function, there is a lack of protocols for designing and tuning the fitness landscapes themselves on which proteins evolve. Here, we introduce the foundations of fitness landscape design (FLD) to quantitatively reshape the biophysical fitness landscape of target proteins. In doing so, we enable near-arbitrary control over the fitness effects of individual protein mutations and even their epistatic interactions. We develop a set of computational FLD algorithms, which use stochastic optimization of an analytically tractable, in vitro-, in silico - , and epidemiologically validated biophysical fitness model to systematically discover antibody ensembles which impose a designed target fitness landscape onto the evolution of a target protein. Theoretical bounds on FLD are derived and validated using a recently published experimental data set of binding affinities between over 62,000 antibody variants and each of three glycoprotein antigens. As an application with broad-ranging possible clinical and public health benefits, we focus on the specific problem of designing antibody ensembles to suppress the fitness trajectories of viral surface protein escape mutations. Our FLD algorithms find that the fitnesses of SARS-CoV-2 genotype neutral networks can be suppressed and that it is possible to discover proactive vaccines that preemptively suppress viral escape mutants’ fitnesses. Our results also support the realizability of laboratory protein evolution experiments with quantitatively programmable fitness landscapes. More broadly, by thinking several steps ahead of pathogen evolution, biophysical FLD opens the door to proactive vaccine, antibody, and peptide design, offering potential solutions to immune evasion of pathogens and cancers as well as improved biosecurity and pandemic preparedness.
Mohanty et al. (Sun,) studied this question.