Identifying Lung Cancer-Associated Genes Using DNA Data

Abstract

Lung cancer remains the leading cause of cancer-related mortality worldwide, with its molecular heterogeneity posing significant challenges for early detection and targeted therapy, which demands personalized treatment approaches and continual genomic monitoring. Fortunately, the latest advances in genomics and machine learning (ML), including deep learning (DL), offer unprecedented opportunities to identify robust biomarkers and better understand disease mechanisms. In this thesis, we focus on applying both ML and DL models, while combining different types of data, as an attempt to identify genes most strongly linked to this type of cancer. We propose a pipeline, implemented in Python, to analyze RNA sequencing data from the TCGA LUAD (Lung Adenocarcinoma) project, combined with important clinical information. Feature selection was carried out using the Random Forest algorithm, and predictive performance was evaluated using multiple classifiers, including Support Vector Machines, Logistic Regression, and Convolutional Neural Networks. Our approach yields promising results by identifying the 20 top-ranked genes, which were cross-referenced with the literature. Several genes were known to be implicated in LUAD, suggesting biological relevance and validating our methodology. Through this work, we highlight the potential of bioinformatics and artificial intelligence in transforming complex genetic data into meaningful insights, with promising implications for early diagnosis and personalized treatment strategies.