A Computational Systems Approach to Elucidate New Mechanisms Involved in Progressive Lung Disease

Abstract

Mucosal surfaces in the lung interface with the outside environment for breathing purposes, but also provide the first line of defense against invading pathogens. The intricate balance of effective immune protection at the pulmonary epithelium without problematic inflammation is not well understood, but is an important consideration in complex lung diseases such as idiopathic pulmonary fibrosis (IPF) and chronic obstructive pulmonary disease (COPD). Although IPF is a fibrotic interstitial lung disease of unknown origin and COPD is an obstructive lung disease, they do share some similarities. Both are heterogeneous and progressive in nature, have no cure and few treatment options, advance through unknown mechanisms, and involve an aberrant immune response. As research has focused into the role the immune system plays in IPF and COPD, it has become clear that disease progression is caused by a complex dysregulation of immune factors and cells across the tissue compartments of the lungs and blood. Data-driven modeling approaches offer the opportunity to infer protein interaction networks, which are able to identify diagnostic and prognostic biomarkers and also serve as the basis for new insight into systems-level mechanisms that define a disease state. Additionally, these approaches are able to integrate data from across multiple tissue compartments, allowing for a more holistic picture of a disease to be formed. Here, we have applied data-driven modeling approaches including partial least squares discriminant analysis, principal component analysis, decision tree analysis, and hierarchical clustering to high-throughput cell and cytokine measurements from human blood and lung samples to gain systems-level insight into IPF and COPD. Overall we found that these approaches were useful for identifying signatures of proteins that differentiated disease state and progression better than current classifiers. We also found that integrating protein and cell measurements across tissue compartments generally improved classification and was useful for generating new mechanistic insight into progression and exacerbation events. In evaluating IPF progression, we showed that the blood proteome of progressors, but not of non-progressors, changes over time, and that our data-driven modeling techniques were able to capture these changes. Curiously, our models showed that complement system components may be associated with both COPD disease state and IPF disease progression. Lastly, though our analysis suggested that circulating blood cytokines were not useful for differentiating disease state or progression, preliminary work suggested that cell-cell communication networks arising from stimulated peripheral blood proteins may be more useful for classification and gaining mechanistic insight from minimally invasive blood samples. Overall, we believe that this approach will be useful for studying the mucosal immune response present in other diseases that are also progressive or heterogeneous in nature.