Host transcriptional meta-signatures reveal diagnostic biomarkers for Plasmodium falciparum malaria.

Abstract

Transcriptomics has been used to evaluate immune responses during malaria in diverse cohorts worldwide. However, the high heterogeneity of cohorts and poor generalization of transcriptional signatures reported in each study limit their potential clinical applications. We compiled 28 public datasets containing 1,556 whole blood or peripheral blood mononuclear cells (PBMC) transcriptome samples. We estimated effect sizes with Hedges´ g and DerSimonian-Laird random effects model for meta-analyses of uncomplicated malaria. Random forest models identified gene signatures that discriminate malaria from bacterial infections or malaria severity. Parasitological, hematological, immunological, and metabolomics data were used for validation. We identified three gene signatures denominated the uncomplicated Malaria Meta-Signature (uMMS), which discriminates P. falciparum malaria from uninfected controls; the Malaria or Bacteria Signature (MoBS), that distinguishes malaria from sepsis and enteric fever; and the cerebral Malaria Meta-Signature (cMMS), which characterizes individuals with cerebral malaria. These signatures correlate with clinical hallmark features of malaria. Blood transcription modules (BTM) indicate immune regulation by glucocorticoids, whereas cell development and adhesion are associated with cerebral malaria. Transcriptional meta-signatures reflecting immune cell responses provide potential biomarkers for translational innovation and suggest critical roles for metabolic regulators of inflammation during malaria.