Air molecular contaminants (AMC) are one of the major difficulties in air treatment, featuring low concentration but high harm. Catalytic ad-sorption is an effective means to remove AMC, and its process depends to a large extent on the physicochemical properties of the contaminants. Improving the development method of catalytic adsorption materials is of great significance. In this paper, by using a variety of already synthesized single-atom catalyst structures and through a combination of first principles and machine learning, important features in the adsorption process were extracted, and a catalytic adsorption descriptor applied to the O2 oxidation means was developed. The results show that the process of catalytic adsorption is related to the active site itself, the sur-rounding environment, and the characteristics of the system, which is the result of multi-factor coupling. On this basis, the controllable synthesis of the material physical property parameters was achieved through the in-situ template method. Dynamic adsorption tests showed that the descriptor has a strong correlation with the adsorption amount of polluting gases, demonstrating the applicability of the descriptor. By combining theory and experiment to link the microscopic mechanism of the material with the macroscopic physical properties, it is of great significance for the development of catalytic materials.