Figure 6: Network analysis of patient-specific response to infection.
From: Identification of a human neonatal immune-metabolic network associated with bacterial infection

(a) Patient-specific expression profiles. Networks of 824 statistically differentially expressed probes (using eBayes with Benjamini–Hochberg correction adj.P≤0.01, fold change ≥2) visualized using BioLayout Express 3D. Three groups of genes are identified, which correspond to those identified by hierarchical clustering in Fig. 1b. Co-expressed genes within networks were then defined with a Pearson correlation r=0.78 and by applying a Markov clustering (MCL) inflation value of 4 and pre-inflation value of 3 and are shown as coloured clusters (for example, Cluster 01 and so on). (b) Twelve MCL clusters of genes are displayed for bacterial infected patients to show patient-specific responses. Average expression values of MCL clusters were calculated from per gene normalized against mean of control samples and ordered by Euclidean distance. Full expression profiles of these clusters can be seen in Supplementary Fig. 6. (c) Twelve gene expression clusters showing heterogeneity within the group of infected neonates were tested for association with clinical parameters. For each infected neonate and cluster, median gene expression was calculated and tested for statistical association with each of the clinical parameters (see Supplementary Data set 4). Wilcoxon rank sum tests were used to test association with binary clinical parameters, Spearman rank correlation tests were used to test association with continuous clinical parameters (sugar, neutrophil count, postnatal age, platelets, duration on antibiotics, pH, white cell count (WCC), hemoglobin (Hb), gestational age). The colour scale represents the P value estimate and asterisks identify Wilcoxon rank sum or Spearman rank correlation significant test outcome at P≤0.05. (d) Cell type enrichment analysis performed at an individual patient level shows patient-specific responses. Median signal levels for cell type-specific gene lists as detailed in Supplementary Data set 2 are plotted (±median absolute deviation) for B, T, NK cells, neutrophil, monocyte and dendritic cells. Control, blue; bacterial infected, red.