The paper introduces mCSM (Mutation Cutoff Scanning Matrix), a novel approach to predict the effects of mutations in proteins using graph-based signatures. These signatures encode distance patterns between atoms, representing the protein residue environment. The method evaluates the impact of mutations on protein stability and interactions, including protein-protein and protein-nucleic acid interactions. mCSM is evaluated on various datasets, showing performance comparable to or better than other widely used methods. It successfully predicts stability changes in the tumor suppressor protein p53, demonstrating its applicability in disease-related mutations. The mCSM approach is based on machine learning and does not require homology models, making it a versatile tool for studying the effects of mutations.The paper introduces mCSM (Mutation Cutoff Scanning Matrix), a novel approach to predict the effects of mutations in proteins using graph-based signatures. These signatures encode distance patterns between atoms, representing the protein residue environment. The method evaluates the impact of mutations on protein stability and interactions, including protein-protein and protein-nucleic acid interactions. mCSM is evaluated on various datasets, showing performance comparable to or better than other widely used methods. It successfully predicts stability changes in the tumor suppressor protein p53, demonstrating its applicability in disease-related mutations. The mCSM approach is based on machine learning and does not require homology models, making it a versatile tool for studying the effects of mutations.