Kolmogorov-Arnold Network Model Integrated with Hypoxia Risk for Predicting PD-L1 Inhibitor Responses in...

Area of Science:

• Oncology
• Bioinformatics
• Artificial Intelligence

Background:

• Hepatocellular carcinoma (HCC) is a major cause of cancer mortality.
• Immunotherapy is a primary treatment for advanced HCC.
• Tumor hypoxia significantly impacts HCC progression and treatment resistance.

Purpose of the Study:

• To develop and validate an artificial intelligence (AI) model for predicting immunotherapy response in HCC.
• To identify hypoxia-related genomic factors influencing immunotherapy outcomes.
• To integrate hypoxia and genomic data for improved predictive accuracy.

Main Methods:

• Utilized publicly available genomic datasets (TCGA-LIHC, GSE233802, EGAD00001008128).
• Selected hypoxia-related genes (HCC-Hypoxia Overlap and immunotherapy response to hypoxia genes) using differential expression and enrichment analyses.
• Employed Synthetic Minority Over-sampling Technique for class balancing.
• Developed a Kolmogorov-Arnold Network (KAN) model and integrated it with a hypoxia model using Support Vector Machine (SVM).

Main Results:

• A hypoxia model was built using 10 genes.
• The KAN model with 11 genes achieved 70% test accuracy.
• The integrated SVM model combining hypoxia and KAN achieved 72.5% test accuracy.
• The AI model effectively predicted immunotherapy response based on hypoxia risk and genomic factors.

Conclusions:

• An AI model integrating hypoxia and genomic data can predict immunotherapy response in HCC.
• This model may help identify potentially treatable genomic factors in HCC patients.
• The findings support the role of hypoxia in HCC immunotherapy response and offer a novel predictive tool.