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

Area of Science:

• Oncology
• Bioinformatics
• Machine Learning

Background:

• Non-small cell lung cancer (NSCLC) is the most common lung cancer subtype.
• Accurate diagnosis of NSCLC subtypes like adenocarcinoma and squamous cell carcinoma is challenging with conventional methods.
• Existing diagnostic techniques can be slow and inconclusive.

Purpose of the Study:

• To develop an accurate computational model for NSCLC subtype classification.
• To address the limitations of conventional diagnostic methods for lung cancer.

Main Methods:

• Proposed RPSLearner, combining random projection (RP) for dimensionality reduction and stacking ensemble learning.
• Utilized multiple independent RP matrices to reduce dimensionality of RNA-seq data.
• Employed a stack of diverse base classifiers with predictions integrated via a deep linear layer network.

Main Results:

• RPSLearner outperformed state-of-the-art approaches in lung cancer subtype classification on 1,333 NSCLC patients.
• Demonstrated efficient preservation of sample-to-sample distances post-dimension reduction.
• The meta-model and feature fusion method showed superior performance compared to individual base models and conventional score ensemble methods.

Conclusions:

• RPSLearner is a promising model for clinical diagnosis of lung cancer subtypes.
• The model's potential for extension to other cancer subtyping applications was highlighted.