AI-Guided Chemotherapy Optimization in Lung Cancer Using Genomic and Survival Data.

Area of Science:

• Oncology
• Bioinformatics
• Computational Biology

Background:

• Adjuvant chemotherapy (ACT) improves survival in early-stage non-small cell lung cancer (NSCLC), but patient benefit is highly variable.
• Identifying suitable candidates for ACT is a significant challenge in precision oncology.

Purpose of the Study:

• To develop and compare machine learning models for predicting survival and recommending ACT versus observation (OBS) in early-stage NSCLC patients.
• To identify genomic features associated with treatment response.

Main Methods:

• Constructed a meta-database from two public NSCLC gene expression datasets (GSE37745, GSE29013).
• Performed feature selection using Cox-based univariate screening with leave-one-out cross-validation.
• Developed and compared three survival models: bagging with elastic net penalized Cox regression, Random Survival Forests (RSF), and DeepSurv neural survival networks.

Main Results:

• RSF achieved the highest predictive performance (test C-index = 0.885).
• Model-based recommendations improved survival in both training and test datasets (Kaplan-Meier analysis).
• Identified key genomic features (TTR, MTURN, ETV3) for treatment response stratification.

Conclusions:

• Machine learning survival models, particularly RSF, effectively identify NSCLC patients who benefit from ACT.
• This data-driven approach supports individualized chemotherapy decisions for early-stage NSCLC.
• Contributes to advancing personalized treatment strategies in NSCLC.