Interactions between olfactory receptors (ORs) and specific odorant molecules encode many distinct odors through intricate activation patterns. In this study, in order to enhance our understanding of olfactory perception, Atomevo-Odor (http://cslodordatabase.7fx.cn/), a comprehensive database for odorants, ORs, and high-quality OR-odorant responses combining experimental data and artificial intelligence prediction, was constructed. Moreover, graph theory and unsupervised learning methods were employed to classify the odorants, and the relationship between odorant functional groups and fragrance types was examined, along with the recognition mechanism of ORs for different odorant functional groups. Furthermore, a CNN-based model was developed for the OR-odorant response prediction. Finally, predictions of unseen data facilitated the identification of potentially responsive OR-odorant pairs, which allowed for further analysis of the response and recognition mechanisms of odorants by ORs. This study provides valuable insights into the design and guidance for subsequent experiments.