A Single Structure-Derived Computational Metric Predicts High-Affinity Antibody Selection Against a Malaria Antigen

There is an increasing need for improved malaria antibodies that can be used in passive immunization strategies to reduce the burden of malaria in endemic regions. Despite considerable progress, the identification or development of variants that meet stringent performance requirements remains a challenge. A key strategy has been the improvement of prototypic antibodies targeting the repeat antigens on Plasmodium falciparum circumsporozoite protein (PfCSP). In this work, we derive a computational metric from predicted protein structures that efficiently captures affinity information of antibody variants of the PfCSP-targeting antibody, CIS43. We then use this metric to rapidly explore sequence space as large as >3×10^47 variants using principles of the germinal center, deriving new high-affinity CIS43 variants from the method. We further extend this framework to generate high-affinity variants of an unrelated PfCSP-targeting antibody, L9, by maturing both homotypic and antigen-binding interactions, which demonstrates substantial flexibility of the approach. Taken together, we show that coupling micro-evolutionarily selected mutations to in silico screening permits the selection of high-affinity malaria antibodies.