p53 is one of the most frequently mutated genes in human cancer and is mutated in the early stages of lung, skin, head, and neck, and esophageal cancers. Unlike other tumor suppressors, most cancer-associated mutations in p53 are missense mutations residing in its DNA-binding domain that lead to loss of tumor suppressor activity and/or gain of novel oncogenic functions. However, it can be challenging to tease apart the molecular consequences of a specific mutation, and even to identify whether a specific mutation is a cancer driver or relatively benign passenger variant. To address this, we used a structure-based approach to build an empirical tool which could characterize the molecular consequences of p53 mutations. Using an extensive experimental mutational dataset of 2967 unique non-synonymous missense mutations; 17 in-silico structural and functional measurements were calculated for every variant using the p53 crystal structure. K-Nearest Neighbor algorithm (lazy IBK) was used to build the classifier which could accurately identify cancer driver and passenger variants (91% accuracy). A blind test set of 996 Cosmic mutations had a performance accuracy of 89.5%. This approach gives us the information on different structural and functional effects of the variants, therefore the tool can predict the molecular stabilizers that are being developed/in trials work best on destabilizing variants, not variants that affect the conformational flexibility of P53. This highlights how structural information can be used to identify and characterize functional variants of p53, guiding personalized medicine and the development of new therapies.