Gene dependency (GD) prediction has become a critical tool in cancer research, aiding in the identification of therapeutic targets and providing insights into tumor biology. In this paper, we present a domain adaptation framework for predicting GD scores across lung cancer subtypes (e.g., adenocarcinoma, mesothelioma, squamous cell carcinoma) using CRISPR-Cas9 and ribonucleic acid (RNA) interference data adapted to The Cancer Genome Atlas (TCGA) patient tumor data. Our framework leverages a deep learning model to transfer knowledge from cell line data (source domain) to patient tumor data (target domain), enhancing generalization across different environmental settings for cancer cells. The model is evaluated using metrics such as mean absolute error (MAE), Pearson correlation, and \(R^2\) , demonstrating high accuracy and robustness on both source and target domains. These findings underscore the value of GD scores in capturing molecular distinctions across lung cancer subtypes and inform strategies for personalized treatment.