Decision tree-based classifiers for lung cancer diagnosis and subtyping using TCGA miRNA expression data

Area of Science:

• Oncology
• Bioinformatics
• Genomics

Background:

• Lung cancer exhibits the highest global cancer mortality rate, underscoring the urgent need for effective biomarker discovery.
• While differential expression analysis is common, machine learning approaches for identifying molecular biomarkers in lung cancer are less explored.
• MicroRNAs (miRNAs) are increasingly recognized for their role in cancer development and progression.

Purpose of the Study:

• To apply machine learning techniques to The Cancer Genome Atlas (TCGA) lung cancer datasets for molecular profiling.
• To develop classification models using specific microRNAs for diagnosing lung cancer and differentiating between major subtypes.
• To identify novel miRNA-based biomarkers for lung adenocarcinoma (LUAD) and lung squamous cell carcinoma (LUSC).

Main Methods:

• Utilized The Cancer Genome Atlas (TCGA) datasets for lung adenocarcinoma (LUAD) and lung squamous cell carcinoma (LUSC).
• Employed empirical negative control microRNAs (miRs) for normalization of TCGA datasets.
• Developed decision tree models to classify lung cancer status and subtype using specific miRNAs.

Main Results:

• Two primary classification models were built using four miRNAs for lung cancer diagnosis and subtyping.
• hsa-miR-183 and hsa-miR-135b effectively distinguished lung tumors from adjacent normal tissues.
• hsa-miR-944 and hsa-miR-205 were identified for classifying tumors into LUAD and LUSC subtypes, with subtype-specific models also developed.

Conclusions:

• Machine learning models utilizing specific miRNAs can accurately diagnose lung cancer and differentiate between LUAD and LUSC subtypes.
• The identified miRNAs (hsa-miR-183, hsa-miR-135b, hsa-miR-944, hsa-miR-205) show potential as diagnostic and subtyping biomarkers.
• This study highlights the utility of machine learning and miRNA profiling for advancing lung cancer biomarker discovery.