Summary: Computational methods have traditionally struggled to predict the effect of mutations
in antibody-antigen complexes on binding affinity. This has limited their usefulness during antibody
engineering and development, and their ability to predict biologically relevant escape mutations.
Here we demonstrate that graph-based signatures can be used to accurately predict the effect of
mutations on antibody binding affinity. We show that mCSM-AB performs better than comparable methods
that have been previously used for antibody engineering.