Pulmonary hypertension (PH) is a frequent complication in patients with pulmonary fibrosis (PF), whereas the mechanism was not well-understood. This study aimed to explore the influence of immune cell infiltration on PH status based on the genomic expression profiles. Microarray data of GSE24988 were downloaded from the GEO database, including 116 lung tissue samples derived from PF patients with various PH status. Proportion of infiltrated immune cells was evaluated using CIBERSORT, a gene expression-based de-convolution algorithm. A random forest classifier was constructed and out of bag (OOB) cross-validation was carried out for PH prediction. The proportions of immune infiltration cells varied differently in PH samples except T regulatory cells (p-value = 0). Compared with non-PH samples, increased number of naive B cells and plasma cells were identified in PH samples, whereas activated dendritic cells and M2 macrophages were relatively lower (p < 0.05). In the random forest model, these four types of immune cells obtained a higher variable importance score than other cells, including mean decreased accuracy and mean decreased gini evaluation. We ran the OOB cross-validation in each sample of datasets (training set and testing set) and obtained 79 and 69% accuracy, respectively. Abnormal proportions of four types of immune cells were identified in PH samples compared with non-PH samples, suggesting their involvement in PH development. In summary, the immune cell infiltration in PF patients is associated with the PH status of patients, which deserves further investigation in the future.
Keywords: immune cell infiltration; prediction; pulmonary fibrosis; pulmonary hypertension; random forest classifier.
Copyright © 2021 Zhu, Zuo, Hu, Wang, Yang, Qi and Feng.