Sequence statistics reliably predict stabilizing mutations in a protein domain

Abstract

Immunoglobulin variable domains are generally thought of as well conserved platforms providing the base for antigen binding loops of highly varying sequence and structure. However, domain evolution must ensure a balance between optimizing antigen affinity and the requirements of a stable, cooperatively folding domain. Since random mutations can carry a significant penalty for domain stability, constraints are imposed both on the repertoire of germline sequences and on somatic amino acid replacements during affinity maturation. Analyzing these constraints in the conceptual framework of statistical mechanics, we have been able to predict stabilizing mutations in the McPC603 V kappa domain from sequence information alone with better than 60% success rate. The validity of this concept not only has far reaching implications for antibody engineering but may also be generalized to engineer other proteins for higher stability.