Cell-informed host-response profiling distinguishes neutrophil-like and PLA2G7-associated immunometabolic axes in systemic infection

Background Whole-blood host-response signatures can distinguish infection-positive from infection-negative systemic inflammation, but their signal combines changes in leucocyte composition with regulation within cells. We asked whether a retrospectively selected and subsequently fixed blood transcriptomic framework could relate candidate transcripts to granulocyte-associated and monocyte/macrophage-associated expression contexts rather than serve as another stand-alone diagnostic product. Methods A 12-gene framework was retrospectively derived from four contrasts across three studies (444 contrast-level observations; 443 unique arrays) and subsequently fixed for external evaluation using a single-sample transcriptome-rank score. Performance was examined separately across cohorts with different specimens, age groups and comparators. Transcript-score sensitivity analyses, donor-level whole-blood single-cell pseudobulk and cell-line qRT-PCR were used to assess candidate localization and cell-line expression-context concordance. Cis-eQTL Mendelian randomisation and colocalisation were included as exploratory genetic context. Results In the closest target-context cohort, GSE134347, the fixed score achieved an AUROC of 0.827 (95% CI 0.764–0.889) for sepsis versus non-infectious critical illness. Performance was lower in pediatric day-1 SIRS comparisons, indicating dependence on population and comparator context. Adjustment for alternative neutrophil-associated transcript scores produced a marker-dependent ordering, with SMPDL3A among the most attenuated and PLA2G7 among the least attenuated candidates; no candidate remained FDR-significant in the primary six-gene family. Donor-level single-cell analysis and qRT-PCR supported a monocyte/macrophage context for PLA2G7 and granulocyte-like localization of the diagnostic immune markers. Exploratory Mendelian randomisation did not support a susceptibility effect for the testable immunometabolism-associated candidates, and the HLA-DRB1 regional signal lacked evidence of a shared variant under single-variant colocalisation assumptions. Conclusion This study provides an auditable, cell-informed framework that relates granulocyte-associated transcription to PLA2G7-associated monocyte/macrophage localization within the whole-blood infection response. The framework is intended for biological prioritisation, and it defines testable candidates for future studies.