Development of a long noncoding RNA-based machine learning model to predict COVID-19 in-hospital mortality.

Area of Science:

• Biomedical Informatics
• Genomics
• Computational Biology

Background:

• Predictive tools for COVID-19 outcomes are crucial for personalized healthcare and reducing disease burden.
• Machine learning models offer potential for accurate prediction of severe outcomes in SARS-CoV-2 infected patients.

Purpose of the Study:

• To develop and validate a machine learning model for predicting in-hospital mortality risk in COVID-19 patients.
• To identify key predictive features, including genetic markers, for COVID-19 mortality.

Main Methods:

• Analysis of blood samples and clinical data from 1286 COVID-19 patients across European and Canadian cohorts (2020-2023).
• Profiling of 2906 long non-coding RNAs (lncRNAs) using targeted sequencing.
• Development of a feedforward neural network classifier using age and lncRNA LEF1-AS1 as predictive features.

Main Results:

• The model, utilizing age and LEF1-AS1, achieved an AUC of 0.83 and balanced accuracy of 0.78 in a discovery cohort.
• Independent validation in a Canadian cohort confirmed consistent predictive performance.
• Higher LEF1-AS1 levels were associated with a reduced mortality risk (HR 0.54).

Conclusions:

• A novel predictive model incorporating LEF1-AS1 demonstrates significant potential for enhancing COVID-19 patient management.
• LEF1-AS1 is a promising biomarker for assessing mortality risk in SARS-CoV-2 infection.
• The model's components are adaptable for measurement in clinical hospital settings.