An index for deviation distance among amino acid sequences of antibody complementarity determining regions.

Monoclonal antibodies are currently essential biological molecules for immunotherapy and diagnosis. The adaptability of an antibody to humans is crucial for ensuring safety and achieving high therapeutic efficacy as a medicinal drug. A considerable amount of effort has been devoted to predicting the human likeness of recombinant antibody molecules to assess their suitability for clinical use. Many previous studies have utilized a database of antibody sequences. However, a rapid assessment of adaptability is also helpful in the early stage of high-affinity antibody molecular design to a target antigen. To characterize the amino acid sequences of human antibodies, we statistically analyzed 742 antigen-antibody complex structures extracted from the Protein Data Bank. The frequency and position of the appearances of the respective residues are surveyed, and their probabilities were obtained for three complementarity-determining regions of heavy and light chains. In particular, the populations were examined from the viewpoint of which positions the respective residues were likely to appear in each complementary determining region. Based on a statistical analysis, an arithmetic index was proposed to evaluate sequence compatibility with humans and assess antibody models in computational molecular design. An examination using a collection of neutralizing antibodies for viral infectious diseases suggested that the index can distinguish human antibodies from those of mice. An examination of a collection of approved antibody drugs showed a certain degree of correlation between the calculated index and the immunogenicity of the antibody drugs. These evaluations demonstrated the feasibility of the proposed index as a rapid method for evaluating molecular adaptability to medicinal antibodies.