Genetics
Machine learning flags CETP as an immune biomarker for diabetes-tuberculosis comorbidity, AUC 0.804 (Microorganisms 2025)
Original title: Development and Validation of Early Alert Model for Diabetes Mellitus-Tuberculosis Comorbidity
A bioinformatics study mined transcriptome data from diabetes mellitus patients with and without tuberculosis comorbidity (DM-TB), applying differential expression analysis, weighted gene co-expression network analysis and ten machine learning algorithms to identify immune biomarkers. Among 1,090 differentially expressed genes, the authors pinpointed CETP (AUC 0.804, CI 0.744 to 0.864), TYROBP (AUC 0.810) and SECTM1 (AUC 0.811) as immune-related biomarkers for DM-TB. A three-gene early alert model built from these markers reached an AUC of 0.86 (CI 0.813 to 0.907), with a sensitivity of 0.80829 and a specificity of 0.75758, and validated to an AUC of 0.901 on an independent dataset. Pathway analysis implicated NF-kB and MAPK signalling in DM-TB pathogenesis. CETP appears here as one component of a machine-learning biomarker panel for an unrelated infectious-disease comorbidity, not as a target of CETP inhibition.
Original abstract
Diabetes mellitus (DM) and tuberculosis (TB) are two global health challenges that significantly impact population health, with DM increasing susceptibility to TB infections. However, early risk prediction methods for DM patients complicated with TB (DM-TB) are lacking. This study mined transcriptome data of DM-TB patients from the GEO database (GSE181143 and GSE114192) and used differential analysis, weighted gene co-expression network analysis (WGCNA), intersecting immune databases, combined with ten machine learning algorithms, to identify immune biomarkers associated with DM-TB. An early alert model for DM-TB was constructed based on the identified core differentially expressed genes (DEGs) and validated through a prospective cohort study and reverse transcription quantitative real-time polymerase chain reaction (RT-qPCR) for gene expression levels. Furthermore, we performed a detailed immune status analysis of DM-TB patients using the CIBERSORT algorithm. We identified 1090 DEGs associated with DM-TB and further pinpointed CETP (cholesteryl ester transfer protein) (AUC = 0.804, CI: 0.744-0.864), TYROBP (TYRO protein tyrosine kinase binding protein) (AUC = 0.810, CI: 0.752-0.867), and SECTM1 (secreted and transmembrane protein 1) (AUC = 0.811, CI: 0.757-0.864) as immune-related biomarkers for DM-TB patients. An early alert model was developed based on these three genes (AUC = 0.86, CI: 0.813-0.907), with a sensitivity of 0.80829 and a specificity of 0.75758 at a Youden index of 0.56587. External validation using the GSE114192 dataset showed an AUC of 0.901 (CI: 0.847-0.955). Population cohort research and RT-qPCR verified the expression levels of these three genes, demonstrating consistency with trends seen in the training set. KEGG enrichment analysis revealed that NF-κB and MAPK signaling pathways play crucial roles in the DM-TB pathogenic mechanism, and immune infiltration analysis showed significant suppression of certain adaptive immune cells and activation of inflammatory cells in DM-TB patients. This study identified three potential immune-related biomarkers for DM-TB, and the constructed risk assessment model demonstrated significant predictive efficiency, providing an early screening strategy for DM-TB.
Summary written by cetpinhibition.org from the published abstract; figures as published. Page updated 18 August 2026. Methods.