Deep learning-based model for predicting progression in patients with head and neck squamous cell carcinoma

Area of Science:

• Oncology
• Bioinformatics
• Computational Biology

Background:

• Head and neck squamous cell carcinoma (HNSCC) poses significant challenges in predicting disease progression.
• Integrating multi-omics data offers a promising avenue for developing more accurate prognostic models.

Purpose of the Study:

• To develop a deep learning (DL)-based model for predicting disease progression in HNSCC patients.
• To leverage multi-omics data, including RNA sequencing, miRNA sequencing, and methylation data, for enhanced prediction accuracy.

Main Methods:

• Utilized The Cancer Genome Atlas (TCGA) multi-omics data as input for an autoencoder, a DL technique.
• Constructed a Support Vector Machine (SVM)-based prognosis model for progression-free survival (PFS) using autoencoder outputs.
• Validated the model's predictive performance against alternative methods and conducted differential expression and functional enrichment analyses.

Main Results:

• The DL-based model identified two distinct patient subgroups with significantly different PFS (C-index = 0.73), validated across three confirmation sets.
• The DL model outperformed Principal Component Analysis (PCA) and individual Cox-PH models in accuracy and efficiency.
• Identified 348 differentially expressed genes (DEGs), 23 differentially expressed miRNAs, and 55 differentially methylated genes linked to immune-related pathways.

Conclusions:

• The developed DL-based model is a reliable and robust tool for predicting HNSCC disease progression.
• Uncovered key genes and pathways implicated in HNSCC progression, including immune-related processes.
• The model's utility can facilitate the development of personalized therapies and improve patient outcomes in HNSCC.