Metric to distinguish closely related domain families using sequence information

Abstract

Engineered antibodies are emerging as a promising class of therapeutic biomolecules, as well as having applications in medical research. Knowledge on conserved functional and structural regions within antibody domains is imperative in order to rationally design stable and specific antibodies. Of particular interest for the design of therapeutics are antibody variable and constant domains, which are responsible for antigen binding and immune response. These antibody domains are part of the larger immunoglobulin (Ig) V-class and C-class families, respectively. We find that, although both classes belong to the Ig-fold superfamily, the sets of conserved residue positions and identities differ between these classes. We exploit these evolutionary differences to derive a metric based on sequence positional entropy that distinguishes C-class from V-class sequences utilizing only sequence information. By distinguishing different domain families using sequence information alone, we enable the application of domain-specific design strategies without the need for secondary or tertiary structural information.