Extracting predictors for lung adenocarcinoma based on Granger causality test and stepwise character selection

Area of Science:

• Oncology
• Bioinformatics
• Genomics

Background:

• Lung adenocarcinoma is a leading cause of cancer mortality globally.
• High-throughput expression microarrays generate vast data on gene expression, DNA methylation, and miRNA.
• Identifying reliable hub genes as biomarkers for lung adenocarcinoma remains a challenge.

Purpose of the Study:

• To develop a novel method for identifying minimal yet effective gene predictors for lung adenocarcinoma.
• To screen for hub genes that can serve as accurate biomarkers for discriminating between cancerous and healthy individuals.

Main Methods:

• Analysis of three expression microarrays to construct a multi-interaction network.
• Transformation of the undirected network to a directed network using Granger causality test.
• Application of a stepwise feature selection algorithm to identify gene predictors.

Main Results:

• Six gene predictors were identified: TOP2A, GRK5, SIRT7, MCM7, EGFR, and COL1A2.
• These predictors are all cancer-related and offer a concise set for diagnosis.
• Validation across six independent datasets demonstrated high precision, ranging from 95.3% to 100%.

Conclusions:

• A robust and effective method for extracting gene predictors was successfully developed.
• The identified six genes can serve as effective predictors for diagnosing lung adenocarcinoma.
• This approach highlights the potential of specific gene expression patterns in cancer differentiation.