Consistent PTGDR downregulation has limited clinical specificity for sepsis: a retrospective multicohort transcriptomic analysis

Background : Blood transcriptomic signatures can separate sepsis from health, but cross-cohort consistency does not establish specificity against clinically realistic inflammatory alternatives. We corrected the discovery meta-analysis of a historical sepsis panel and evaluated its surviving signal across cohorts. Methods : GSE57065 and GSE95233 were reprocessed from 231 raw microarray CEL files. Moderated standard errors were combined by restricted maximum-likelihood random-effects meta-analysis with ordinary and modified Hartung–Knapp sensitivities. Four historical genes were evaluated, with a family-wise multiplicity sensitivity. Because both cohorts informed candidate prioritization, reciprocal modelling was classified as development rather than independent validation and reported by held-out cohort. Metrics were recomputed from stored prediction records for nine contrasts. A PTGDR-only expression extension specified before outcome inspection evaluated bulk RNA-seq, donor-level single-cell data, covariates and composition. Results : PTGDR was downregulated (pooled effect −1.972, conventional 95% CI −2.326 to −1.619; ordinary Hartung–Knapp P = 0.025), but the modified Hartung–Knapp interval crossed zero (−4.264 to 0.320; P = 0.058) and the four-hypothesis Holm-adjusted P was 0.100. In reciprocal development folds, PTGDR AUC was 0.896 and 0.972 versus 0.540 and 0.520 for the historical panel. External prediction-record AUC was 0.873 for healthy versus infection but 0.554 for infection versus sepsis; against post-surgical inflammation AUC was 0.790, yet fixed-threshold specificity was 0.032. In newly analysed cohorts, sepsis-versus-healthy AUCs were 0.901–0.970, whereas AUCs were 0.582 versus adult ICU controls and 0.589 versus severe COVID-19 ICU patients. A three-cohort synthesis had I² = 67.4% and a prediction interval crossing zero. In 133-donor single-cell data, adult classical-monocyte AUC was 0.831 versus health but 0.630 versus ICU controls. Conclusions : Correcting the historical uncertainty calculation rejected the three-gene panel. PTGDR remained the strongest exploratory signal across cohorts, probes, platforms, and selected immune-cell compartments, but did not show sepsis-specific diagnosis, calibration, or fixed-threshold transportability. Biomarker consistency must therefore be distinguished from clinical utility.

Authors

Institutions

Publication Details

Journal
Research Square
Published
2026-11-18
DOI
https://doi.org/10.21203/rs.3.rs-11576550/v1
Primary Topic
Sepsis Diagnosis and Treatment
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Consistent PTGDR downregulation has limited clinical specificity for sepsis: a retrospective multicohort transcriptomic analysis

Hemantkumar Patadia, Ben Tweve, Yellapu Sampreeth Naidu, Nikhilesh Dangi
4 links
Research Square
Sepsis Diagnosis and Treatment
preprint

Consistent PTGDR downregulation has limited clinical specificity for sepsis: a retrospective multicohort transcriptomic analysis

Hemantkumar Patadia, Ben Tweve, Yellapu Sampreeth Naidu, Nikhilesh Dangi
preprint en

Abstract

Background : Blood transcriptomic signatures can separate sepsis from health, but cross-cohort consistency does not establish specificity against clinically realistic inflammatory alternatives. We corrected the discovery meta-analysis of a historical sepsis panel and evaluated its surviving signal across cohorts. Methods : GSE57065 and GSE95233 were reprocessed from 231 raw microarray CEL files. Moderated standard errors were combined by restricted maximum-likelihood random-effects meta-analysis with ordinary and modified Hartung–Knapp sensitivities. Four historical genes were evaluated, with a family-wise multiplicity sensitivity. Because both cohorts informed candidate prioritization, reciprocal modelling was classified as development rather than independent validation and reported by held-out cohort. Metrics were recomputed from stored prediction records for nine contrasts. A PTGDR-only expression extension specified before outcome inspection evaluated bulk RNA-seq, donor-level single-cell data, covariates and composition. Results : PTGDR was downregulated (pooled effect −1.972, conventional 95% CI −2.326 to −1.619; ordinary Hartung–Knapp P = 0.025), but the modified Hartung–Knapp interval crossed zero (−4.264 to 0.320; P = 0.058) and the four-hypothesis Holm-adjusted P was 0.100. In reciprocal development folds, PTGDR AUC was 0.896 and 0.972 versus 0.540 and 0.520 for the historical panel. External prediction-record AUC was 0.873 for healthy versus infection but 0.554 for infection versus sepsis; against post-surgical inflammation AUC was 0.790, yet fixed-threshold specificity was 0.032. In newly analysed cohorts, sepsis-versus-healthy AUCs were 0.901–0.970, whereas AUCs were 0.582 versus adult ICU controls and 0.589 versus severe COVID-19 ICU patients. A three-cohort synthesis had I² = 67.4% and a prediction interval crossing zero. In 133-donor single-cell data, adult classical-monocyte AUC was 0.831 versus health but 0.630 versus ICU controls. Conclusions : Correcting the historical uncertainty calculation rejected the three-gene panel. PTGDR remained the strongest exploratory signal across cohorts, probes, platforms, and selected immune-cell compartments, but did not show sepsis-specific diagnosis, calibration, or fixed-threshold transportability. Biomarker consistency must therefore be distinguished from clinical utility.

Research Square
Parul University (IN)
Good health and well-being
Sepsis Diagnosis and Treatment
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon