Circulating lncRNAs as biomarkers for severe dengue using a machine learning approach

Area of Science:

• Molecular Biology
• Infectious Diseases
• Biomarker Discovery

Background:

• Dengue virus (DENV) infection poses a global health threat, with severe dengue (SD) causing significant mortality.
• Current methods for predicting SD progression lack specificity, straining healthcare systems in endemic areas.
• Circulating long non-coding RNAs (lncRNAs) show potential as stable and reliable biomarkers.

Purpose of the Study:

• To investigate the utility of lncRNAs as predictive markers for severe dengue (SD).
• To develop a predictive model for SD progression using lncRNAs.
• To assess the clinical applicability of lncRNA-based prediction in dengue patient management.

Main Methods:

• Differential gene expression analysis and quantitative PCR (qPCR) arrays were used to identify candidate lncRNAs.
• Plasma lncRNA levels were quantified in Vietnamese dengue patients (n=377) and healthy controls (n=128).
• Machine learning algorithms were employed to build a predictive model for SD risk.

Main Results:

• The developed lncRNA predictive model achieved high sensitivity and specificity (AUC = 0.98).
• The model accurately identified 17 out of 18 non-surviving patients as high-risk upon admission.
• This significantly outperformed traditional warning signs, which identified only 9 of these critical cases.

Conclusions:

• A panel of lncRNAs demonstrates significant potential for predicting severe dengue (SD).
• This biomarker approach could enhance early patient stratification and management in dengue-endemic regions.
• Implementing lncRNA-based prediction may alleviate healthcare system burdens and improve patient outcomes.