Classify multicategory outcome in patients with lung adenocarcinoma using clinical, transcriptomic and...

Area of Science:

• Oncology
• Bioinformatics
• Computational Biology

Background:

• Accurate classification of multi-category survival outcomes is crucial for personalized cancer treatment strategies.
• Machine learning (ML) has shown promise in predicting survival outcomes for various cancer types, but its application to lung adenocarcinoma is limited.
• Lung adenocarcinoma (LUAD) survival outcome classification requires robust predictive models.

Purpose of the Study:

• To compare the performance of machine learning and statistical models in classifying four-category survival outcomes for lung adenocarcinoma.
• To identify the most effective data types (clinical, transcriptomic, or combined) for accurate LUAD survival outcome prediction.
• To discover novel gene biomarkers associated with specific LUAD survival outcomes.

Main Methods:

• Utilized The Cancer Genome Atlas (TCGA) dataset for lung adenocarcinoma.
• Compared four models: random forests, support vector machine (SVM), multilayer perceptron, and multinomial logistic regression (Mlogit).
• Evaluated model performance using micro-average area under the curve (AUC) and analyzed feature importance (genes).

Main Results:

• Multinomial logistic regression (Mlogit) and support vector machine (SVM) models showed similar performance (micro-average AUC=0.82), outperforming random forests.
• Transcriptomic data alone and combined clinico-transcriptomic data were sufficient for accurate classification, unlike clinical data alone.
• Identified specific gene signatures associated with different survival outcomes (e.g., alive without disease, alive with progression, dead with disease).

Conclusions:

• Machine learning and Mlogit models can effectively classify lung adenocarcinoma survival outcomes.
• Transcriptomic data is a key predictor of LUAD survival, highlighting its importance in precision oncology.
• Identified gene signatures offer potential for risk stratification and development of targeted therapies for lung adenocarcinoma.