Computational immunology has been the breeding ground of some of the best bioinformatics work of the day. By melding diverse data types, these approaches have been successful in associating genotypes with phenotypes. However, the representations (or spaces) in which these associations are mapped have primarily been constructed from some omics-oriented sequence data typically derived from high-throughput experiments. In this perspective, we highlight the importance of biophysical representations for performing the genotype–phenotype map. We contend that using biophysical representations reduces the dimensionality of a search problem, dramatically expedites the algorithm, and more importantly, offers physical interpretability to the classes of clustered sequences across different layers of complexity – molecular, cellular, or macro-level. Such biophysical interpretations offer a firm basis for the future of bioengineering and cell-based therapies.
