In current drug discovery, a high percentage of Phase II clinical trials fail to obtain a proof of concept (POC). To solve this problem, we wanted to establish a method to search for drug targets using patient information. Although the use of AI technology in the medical and drug discovery fields is becoming popular, the clinical information is often not machine-readable, and there was an urgent need to develop a method to fully utilize it. In this study, we selected IPF (idiopathic pulmonary fibrosis), which has an ample room for exploration of effective drugs, as a target disease and developed a strategy to identify drug targets by processing medical records into structured data and further combining them with biomolecular information obtained from the serum exosome proteome. These data were analyzed in a data-driven manner and several proteins were found to be highly expressed in a group of patients with clinical features of IPF. These proteins included molecules targeted by nintedanib, but also included molecules that have not been reported to be associated with IPF. The pathway analysis identified several biological networks and several agents, for example, ponatinib, the multi-tyrosine kinase inhibitor, were predicted as upstream regulators of these networks, and their inhibitory effects on epithelial-mesenchymal transition (EMT) were confirmed. In this study, we have developed a strategy to use real-world clinical information, and we believe that this strategy can be widely applied to a variety of diseases and help the searching of drug targets in a sustainable way.