Figure 5. Investigating I₀.

Examples of approaches to identify characteristic features of I₀ clusters (A–C) and infer I₀ associations with susceptibility to infection or disease severity (D). (A) Example of overlaying the value of a putative independent variable, cytomegalovirus (CMV) status, on a principal component analysis (PCA) graph. (B) Example of subsetting the data by cluster identity and comparing values of a potential defining variable, age. (C) Example of a PCA loading plot. (D) Example of a decision tree analysis, in which an I₀ value, PC₁, is associated with susceptibility to severe disease. Green lines = true, yellow lines = false. *Note: Examples are not related to each other.