Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning

Area of Science:

• Oncology
• Molecular Biology
• Immunology

Background:

• Diffuse large B-cell lymphoma (DLBCL) is a heterogeneous non-Hodgkin lymphoma with variable patient prognosis.
• Programmed cell death (PCD) is crucial in cancer development and progression.
• Analyzing PCD-related gene expression in DLBCL can refine risk stratification and personalize treatments.

Purpose of the Study:

• To identify differentially expressed genes (DEGs) related to PCD in DLBCL.
• To classify DLBCL subtypes based on PCD gene expression and immune microenvironment.
• To develop predictive models for DLBCL using machine learning.

Main Methods:

• Integrated five DLBCL datasets with 18 PCD-related gene expression profiles.
• Utilized consensus clustering, immune infiltration analysis, GSVA, and WGCNA.
• Employed 12 machine learning algorithms and transcriptome sequencing for validation.

Main Results:

• Identified 1074 PCD-related DEGs, revealing two distinct DLBCL molecular subtypes.
• The C2 subtype, characterized by upregulated DNA repair and cell cycle pathways, was classified as high-risk.
• Five potential biomarkers (CTSB, DPYD, SCARB2, STOM, GBP1) were identified and validated.

Conclusions:

• Two DLBCL subtypes based on PCD gene expression were identified, with C2 being high-risk.
• The identified biomarkers may enhance DLBCL risk stratification and understanding of its heterogeneity.
• Findings provide a basis for further research into DLBCL progression and prognostic improvements.