In (Taft et al., 2021) machine learning / deep learning (ML/DL) models predict the binding of various mutated variations of SARS-CoV-2 receptor binding motif (RBM) to ACE2 and 4 therapeutic monoclonal antibodies (mAbs). Both training and validation use data with both single and multiple mutations in SARS-CoV-2 RBM. Thus, models solve interpolation problems. Research question: is it possible to use single mutations SARS-CoV-2 RBM for training to predict binding with multiple mutations SARS-CoV-2 RBM (extrapolation problems)?