Computationally generated molecular fragments offer an effective way to represent essential molecular properties, especially when detailed mechanistic information is limited. These fragments provide a means to categorize molecules into different classes by encoding functional groups responsible for specific chemical, physical, toxic, or biological characteristics. In drug discovery, utilizing libraries of small fragments is more efficient than whole molecules due to their reduced number compared to drug-like compounds. Molecular fragments serve both educational and practical purposes, aiding in understanding structure–activity relationships and facilitating the development of interpretable QSAR models for virtual screening and regulatory decision making. Despite their simplicity, the computational representation and handling of molecular fragments involves sophisticated techniques. This chapter describes common concepts, various fragment representations, and illustrates their application in building QSAR/QSPR models through examples.