Building a model of sepsis: data integration unravels pathogenic mechanisms in severeP. aeruginosainfections

Abstract

Understanding how human hosts and bacterial pathogens interact with each other is essential to explaining the differences in severity and outcomes of systemic infectious diseases (sepsis). Such pathogens, likePseudomonas aeruginosa(PA), remain a major public health concern. In this work we present a data integration workflow to build a model of interaction betweenPAand its human host. The pathogenic mechanisms ofPAinfection are described through accurate mapping of protein-protein, and metabolite-protein direct interactions, betweenPAand human host, as well as a description of pathways activated in organs affected by severePAinfections.In the first step, a literature review process was carried out based on previous experiences with virus-host interaction models. The collected Data was organized to build a conceptual model on multiple levels, providing a view of the relevant mechanisms involved in severe bacterial infections. All information aboutPAinfection was categorized into three main groups: a) conceptual information about pathogenesis; b) molecular interactions between bacterium and humans; and c) gene expression signatures and dysregulated pathways in severe infections. A dataset of 89PA– human interactions, involving 152 proteins/molecules (109 human proteins, 3 human molecules, 34PAproteins, and 5PAmolecules) was reviewed and annotated, providing a new perspective on thePA-host interaction to improve understanding of the host biological responses. Such data were complemented by transcriptomic data onPA-infected lung samples, highlighting overexpression of proinflammatory pathways as well as modulation of the activities ofPAon the host response in the lung.ImportanceThe intricate interplay between host and microbial pathogens prompts us to use innovative data sources, combining omics data, clinical patient data, and pathogen features, through exploratory computational methods. A data curation process allowed us to put together a detailed description of the pathogenic mechanisms ofPAin the human host, connecting basic knowledge about clinical aspects, interactions between bacterial and host proteins, and pathways dysregulated duringPAinfection. The results of our work allowed us to start putting together a model ofPAinfection, defining specific steps inPAinfections and providing a significant step forward in understanding the deep interaction betweenPAand the human host in severe infections. This approach could be an effective way to predict specific clinical phenotypes in severe bacterial infections.