Idiopathic pulmonary fibrosis is a progressive and debilitating lung disease with large unmet medical need and few treatment options. We describe an analysis connecting single cell gene expression with bulk gene expression-based subsetting of patient cohorts to identify IPF patient subsets with different underlying pathogenesis and cellular changes. We reproduced earlier findings indicating the existence of two major subsets in IPF and showed that these subsets display different alterations in cellular composition of the lung. We developed classifiers based on the cellular changes in disease to distinguish subsets. Specifically, we showed that one subset of IPF patients had significant increases in gene signature scores for myeloid cells versus a second subset that had significantly increased gene signature scores for ciliated epithelial cells, suggesting a differential pathogenesis among IPF subsets. Ligand-receptor analyses suggested there was a monocyte-macrophage chemoattractant axis (including potentially CCL2-CCR2 and CCL17-CCR4) among the myeloid-enriched IPF subset and a ciliated epithelium-derived chemokine axis (e.g. CCL15) among the ciliated epithelium-enriched IPF subset. We also found that these IPF subsets had differential expression of pirfenidone-responsive genes suggesting that our findings may provide an approach to identify patients with differential responses to pirfenidone and other drugs. We believe this work is an important step towards targeted therapies and biomarkers of response.