Abstract Rationale Red blood cell (RBC) transfusions can trigger variable inflammatory responses among critically ill pediatric patients with multiple organ dysfunction syndrome (MODS). We aimed to identify inflammatory biomarkers exhibiting correlated expression patterns indicative of co-regulated responses to RBC transfusion. To uncover such relationships, we constructed a correlation-based protein-protein interaction network to identify markers with coordinated post-transfusion responses. We compared their expression levels between previously derived hyper- and hypo-inflammatory patient groups identified through latent profiling of baseline inflammatory markers. Methods We analyzed pre- and post- transfusion plasma samples from 146 critically ill children with MODS using a multiplex biomarker panel (NULISA; Alamar Biosciences). Biomarker changes were quantified as log2 fold differences between day 0 (pre-transfusion) and day 1 (post-transfusion) measurements. Latent profiles were derived using pre-transfusion values. Pairwise Spearman correlation coefficients of biomarker changes were computed, and preliminary analyses were performed using hierarchical and k-means clustering. A protein-protein interaction network was constructed in Gephi v0.10.1 using the pairwise correlations, filtered to include co-expression relationships with high correlation (|Spearman Rho|0.5) and significance (FDR p-value =0.05). Network edges were colored by correlation direction and scaled in thickness and color based on correlation strength. Node size reflects the marker’s influence in the network using eigenvector centrality, and colors correspond to community assignments derived from modularity analysis. Pathway enrichment of marker clusters was assessed using STRINGdb and a background list of all inflammatory markers (n = 249). Differences in the expression of significant marker clusters between latent profiles were assessed using the Wilcoxon rank-sum test. Results Hierarchical clustering of pairwise Spearman correlation coefficients identified a cluster of inflammatory markers (n = 28) with strong positive correlations that overlapped largely with the k-means clustering. After filtering for network analysis in Gephi, 175 markers were retained and partitioned into six communities, resulting in a cluster of markers (n = 29, green cluster in Fig. 1) with the strongest inter-marker correlations. There was substantial overlap between clusters identified in hierarchical, k-means, and Gephi analyses. This cluster was significantly enriched for the NOD-like receptor pathway (FDR=0.008), while all other clusters showed no significant enrichment. Wilcoxon rank-sum test showed that the overall expression of markers in this pathway was significantly higher in the hyperinflammatory group (p = 0.047). Conclusions Post-transfusion inflammatory responses in pediatric MODS patients may be shaped by baseline immune states and involve coordinated activation of the NOD-like receptor pathway. These findings suggest that this pathway is a potential target for mitigating post-transfusion inflammatory responses. This abstract is funded by: 1RO1HD092471
Wang et al. (Fri,) studied this question.