A computational mutagenesis methodology utilizing a four-body, knowledge-based, statistical contact potential is applied toward quantifying sequence-structure compatibility changes in bacteriophage T4 lysozyme upon single amino acid replacements. We show that these scalar scores correlate with experimentally measured stability changes to the protein due to the mutations. For each mutant, the approach also generates a vector of environmental perturbations occurring at every position in the protein. Implementation of the random forest algorithm, utilizing 521 experimental T4 lysozyme mutants each represented by its respective perturbation vector, correctly classifies mutants based on the direction of stability change with 88% cross-validation accuracy and 0.70 Matthew’s correlation coefficient while achieving 0.91 area under the receiver operating characteristic curve. Learning curves are presented and reveal the dependence of training set size on model performance. The trained random forest model is used to infer stability changes for all remaining unexplored T4 lysozyme mutants.