A Unique Patient Stratification Method Combined with a Machine Learning Approach Identifies Novel Genetic Susceptibility and Protective Factors for Severe COVID-19 in a Hungarian Population

TL;DR

This study uses a new patient classification method and machine learning to find genetic factors linked to severe COVID-19 in a Hungarian population.

Contribution

A novel patient stratification method combined with machine learning to identify new genetic susceptibility and protective factors for severe COVID-19.

Findings

Identified 877 genes with variants distinguishing severe from non-severe COVID-19 cases.

Categorized genes into susceptibility and protective factors for severe disease outcomes.

Gene-set enrichment analysis revealed key biological pathways involved in disease severity.

Abstract

Intensive research has shown that severe COVID-19 outcomes are influenced by antiviral pathways and immune responses, both shaped by genetic predisposition. In this study, we aimed to identify genetic variants associated with disease severity in a cohort of Hungarian patients. We applied a novel stratification method based on age, disease severity, and clinical background to classify patients by susceptibility to severe COVID-19. Whole-exome sequencing (WES) was performed on 168 individuals, and gene mutation loads were assessed. Using a Random Forest machine learning approach, we identified variants of 877 genes that distinguished between severe and non-severe cases. We further categorized these genes as either susceptibility or protective factors. Gene-set enrichment analysis highlighted the most affected biological pathways. Our findings support the development of personalized diagnostic tools to assess the risk of severe COVID-19 and guide targeted treatment strategies. Our findings further extend the results of previous studies, providing novel insights into the genetic determinants of COVID-19 severity.

Figures