Exploring gene expression heterogeneity in burn wounds through machine learning models: in silico...

Area of Science:

• Biomedical Informatics
• Genomics
• Computational Biology

Background:

• Burn wound healing is complex, with significant variability in patient outcomes.
• Understanding gene expression heterogeneity is crucial for developing targeted therapies.
• Previous studies have not fully elucidated the molecular subtypes of burn injuries.

Purpose of the Study:

• To apply machine learning techniques to identify distinct gene expression patterns in burn wound patients.
• To investigate the relationship between gene expression profiles, patient age, and time since injury.
• To evaluate the performance of various clustering algorithms and a linear discriminant analysis model.

Main Methods:

• Utilized gene expression data from the Gene Expression Omnibus (GEO) database.
• Applied principal component analysis (PCA) for data reduction.
• Employed K-Means Clustering, Agglomerative Clustering, Spectral Clustering, and Gaussian Mixture Models for cluster analysis.
• Validated cluster performance using the Silhouette Score and employed Linear Discriminant Analysis (LDA) for classification.

Main Results:

• Identified three distinct gene expression clusters among burn patients.
• Found significant differential gene expression between clusters, involving hundreds of genes.
• Observed significant age-related differences between clusters, with Clusters 0 and 1 representing older individuals.
• Achieved 93% test accuracy using LDA for classifying gene expression data.

Conclusions:

• Gene expression heterogeneity exists in burn wounds, potentially influenced by patient age.
• The study successfully identified molecular subtypes of burn injury based on gene expression.
• These findings support the development of early diagnostic tools and personalized therapeutic strategies for burn patients.