Coupling sparse Cox models with clustering of longitudinal transcriptomics data for trauma prognosis.

Area of Science:

• Genomics and Bioinformatics
• Systems Biology
• Clinical Informatics

Background:

• Longitudinal gene expression analysis and survival modeling provide critical biological and clinical insights.
• High-dimensional time series transcriptomics data presents unique challenges for pattern discovery.

Purpose of the Study:

• To develop a novel framework for discovering gene signatures and patterns in high-dimensional time series transcriptomics data.
• To assess the association between identified gene patterns and hospital length of stay in trauma patients.

Main Methods:

• Applied Cox regression with elastic net regularization for initial dimensionality reduction.
• Developed a novel imputation methodology for missing gene expression values.
• Utilized multivariate time series (MTS) clustering to analyze gene expression trajectories over time and stratify patients.
• Validated patient stratification using Kaplan-Meier curves and log-rank tests.

Main Results:

• Identified 22 genes significantly associated with hospital discharge.
• Early-stage gene expression levels accurately predicted length of stay.
• The proposed imputation method ensured data completeness with minimal information loss.
• MTS clustering successfully grouped patients with similar gene trajectories and hospital discharge times, indicating comparable injury responses.

Conclusions:

• The developed framework effectively integrates time-to-event data with longitudinal, high-dimensional transcriptomics.
• Gene expression trajectories strongly correlate with patient recovery, potentially improving trauma patient management.
• The methodology is adaptable for other medical datasets, enhancing clinical decision support systems.