IBPGNET: lung adenocarcinoma recurrence prediction based on neural network interpretability

Area of Science:

• Oncology
• Bioinformatics
• Computational Biology

Background:

• Lung adenocarcinoma (LUAD) is a prevalent lung cancer subtype.
• Early-stage LUAD patients face a significant risk (30-50%) of metastatic recurrence post-surgery.

Purpose of the Study:

• To develop a novel computational framework, Interpretable Biological Pathway Graph Neural Networks (IBPGNET), for predicting LUAD recurrence.
• To elucidate the underlying regulatory mechanisms of LUAD progression and recurrence.
• To integrate multi-omics data for enhanced interpretability in cancer research.

Main Methods:

• Developed IBPGNET, a graph neural network model leveraging pathway hierarchy.
• Integrated diverse omics data for comprehensive analysis.
• Validated IBPGNET performance using 5-fold cross-validation against existing classification methods.

Main Results:

• IBPGNET demonstrated superior performance compared to other classification methods.
• Identified PSMC1 and PSMD11 as genes significantly associated with LUAD recurrence.
• Observed elevated PSMC1 and PSMD11 expression in LUAD cells versus normal cells.
• Knockdown of PSMC1/PSMD11 enhanced afatinib sensitivity and reduced LUAD cell migration, invasion, and proliferation.
• Demonstrated that PSMC1/PSMD11 may regulate therapeutic sensitivity via EGFR expression.

Conclusions:

• IBPGNET is an effective tool for predicting LUAD recurrence and understanding its mechanisms.
• PSMC1 and PSMD11 are potential biomarkers for LUAD recurrence and therapeutic targets.
• Targeting PSMC1 and PSMD11 could offer new therapeutic strategies for LUAD, potentially through modulating EGFR signaling.