XGBoost-Based Feature Learning Method for Mining COVID-19 Novel Diagnostic Markers

Area of Science:

• Genomics
• Infectious Diseases
• Bioinformatics

Background:

• The COVID-19 pandemic, caused by SARS-CoV-2, emerged in late 2019, leading to a global health crisis and significant economic impact.
• Accurate and rapid diagnostic methods are crucial for managing infectious disease outbreaks.

Purpose of the Study:

• To identify novel diagnostic biomarkers for COVID-19 using gene expression data.
• To develop and validate machine learning models for classifying COVID-19 patients.

Main Methods:

• Downloaded and analyzed throat swab gene expression data from COVID-19 positive and negative patients via the Gene Expression Omnibus (GEO) database.
• Employed XGBoost for feature gene selection and constructed various machine learning classifiers (MARS, KNN, SVM, MIL, RF).
• Utilized the Iterative Feature Selection (IFS) method to select the optimal KNN classifier and identified 24 feature genes, validated using Principal Component Analysis (PCA).

Main Results:

• Identified a set of 24 feature genes capable of effectively classifying COVID-19 positive and negative patients.
• The selected genes were significantly enriched in biological functions related to viral transcription and viral gene expression.
• Pathway analysis indicated enrichment in pathways associated with Coronavirus disease-COVID-19.

Conclusions:

• The 24 identified feature genes demonstrate high efficacy in distinguishing between COVID-19 positive and negative individuals.
• These genes hold promise as novel biomarkers for the diagnosis of COVID-19.
• The findings contribute to the development of more effective diagnostic strategies for the pandemic.