HPLC-HRMS and interpretable machine learning decipher serum lipidomic signatures in NSCLC

Area of Science:

• Oncology
• Biochemistry
• Computational Biology

Background:

• Non-small cell lung cancer (NSCLC) is a leading cause of cancer mortality.
• Lack of non-invasive tools for NSCLC detection and risk stratification hinders effective management.
• Lipid metabolic reprogramming in cancer presents opportunities for biomarker discovery.

Purpose of the Study:

• To identify and validate serum lipid biomarkers for non-small cell lung cancer (NSCLC) detection.
• To develop and evaluate a computational model for NSCLC risk stratification using lipid profiles.
• To explore the potential of lipidomics in non-invasive cancer diagnostics.

Main Methods:

• Serum lipid profiling of 40 NSCLC patients and 30 controls using HPLC-HRMS.
• Application of machine learning models (LASSO, SVM, XGBoost, LightGBM) for classification.
• Development of an integrated lipid-clinical model with nomogram for risk estimation.

Main Results:

• NSCLC samples showed decreased glycerophospholipids, sphingolipids, and triacylglycerols.
• The LightGBM model demonstrated superior discrimination in internal validation.
• An integrated lipid-clinical model achieved an AUC of 0.946 with acceptable calibration.

Conclusions:

• Candidate serum lipid markers and an interpretable modeling workflow for NSCLC classification were identified.
• The findings suggest potential for non-invasive NSCLC detection and risk stratification.
• Further external validation in larger, multicenter cohorts is necessary for clinical implementation.