Objective: This study investigates the expression of mitochondrial autophagy-related genes (MRGs) in asthma to establish a novel model for disease prediction, and also identifies asthma subtypes based on the MRGs to figure out the potential molecular targeted drugs. Methods: The data of asthmatic airway samples were obtained from gene expression omnibus data base. Differentially expressed MRGs were screened and validated so as to build a model for disease prediction by using machine learning algorithms. According to the different MRGs expression pattern, two subtypes of asthma were defined, and the biological functions and signaling pathways were investigated by gene ontology (GO) and kyoto encyclopedia of genes and genomes (KEGG) analysis to find out the potential agents through connectivity map database. Results: MRGs expression in asthma patients was significantly increased compared to healthy subjects. Among these genes, TOMM5 was found to be the top differentially expressed MRGs which were up-regulated both in asthma patients and asthmatic mice or primary cell models. In 22 MRGs, seven genes (TOMM5, FUNDC1, TOMM22, SQSTM1, PGAM5, MFN2, and RPS27A) were screened to establish a model for disease prediction for its good performance exhibited by receiver operating characteristic curve assessment in asthma. Through consensus cluster analysis, two subtypes of asthma were classified considering the differences of gene expression and pathway enrichment. The predicted small molecule agents targeting these two subtypes were XMD8-92 and Verrucarin-A respectively. Conclusion: Seven MRGs were confirmed to be the effective molecular markers for asthma prediction, and our findings provide valuable evidences and open a new insight for the development of individualized approaches for asthma management.