Fig. 1. Analytic flow of integrated-omics analysis.

This flowchart presents a brief overview of the main analytical steps in the current study. The steps are shown in order from top to bottom (A to C). For each of 1) host response (transcriptome), 2) microbial composition (metatranscriptome), and 3) microbial function (metatranscriptome) data elements, we individually performed the analysis in steps A and B. Then, we subsequently integrated these omics data in step C. A We examined the relationship of each omics data element with the risk of PPV use at the individual data level. B To reduce the dimensions of the host response, microbial composition, and microbial function data, we performed a weighted gene co-expression network analysis and identified distinct networks (modules). In each omics element, we selected the top five modules with the highest correlation of PPV use and biological significance for the subsequent integrated analyses. C Finally, to determine the integrated relationships of these dual-transcriptome modules with the risk of PPV use, we constructed a logistic regression model with ridge regularization. To uncover the causal relationship structure between these dual-transcriptome modules, we also applied a causal structural learning approach. Abbreviations: FDR, false discovery rate; PPV, positive pressure ventilation; WGCNA, weighted gene co-expression network analysis.