RPSLearner: A Novel Approach Based on Random Projection and Deep Stacking Learning for Categorizing Non-Small Cell Lung Cancer.

Non-small cell lung cancer (NSCLC) comprises the largest subtype of lung cancer with the most cases. Lung adenocarcinoma and lung squamous cell carcinoma are two NSCLC subtypes that pose challenges for accurate diagnosis using conventional methods, including histological examination and imaging, which can be slow and inconclusive. To address these concerns, RPSLearner is proposed, which combines random projection (RP) for dimensionality reduction and stacking ensemble learning to accurately predict lung cancer subtypes. Specifically, multiple independent RP matrices are first generated to project the high-dimensional RNA-seq data into a lower-dimensional space, whose features are subsequently concatenated. After that, the concatenated RP features are fed into a stack of diverse base classifiers, and integrated the predictions from base models via a deep linear layer network. Benchmarking tests on 1 333 NSCLC patients demonstrated that RPSLearner outperformed state-of-the-art approaches for lung cancer subtype classification. Specifically, RPSLearner efficiently preserved sample-to-sample distances even after significant dimension reduction, and the meta-model in RPSLearner yielded consistently higher scores than individual base models. In addition, the feature fusion method outperformed conventional score ensemble methods. We believe RPSLearner is a promising model for downstream lung cancer clinical diagnosis, and it holds the potential to be extended to subtyping of other types of cancer.