<p>Accurately predicting anti-cancer drug responses in cell lines is essential for advancing precision medicine in oncology. Despite the development of various computational models, further improvement is needed, particularly in handling unseen data that is not part of the training set. One major challenge in precision medicine is selecting the best treatment for each patient based on personalized information. To address this issue, a novel method is proposed to predict drug responses by leveraging multiple types of genome-wide molecular data. First, an optimal subset of drugs is identified based on response similarities. Then, coefficients are generated based on the responses of these selected drugs to each cancer cell. The Grey Wolf Optimization (GWO) algorithm is used to assign weights to these coefficients, allowing for more accurate prediction of drug responses. Unlike many existing models, this approach does not require high-end hardware and relies solely on IC50 values of drugs across cell lines for prediction. Experimental results demonstrate superior performance compared to state-of-the-art models, particularly when applied to large datasets that include unseen data. The proposed method achieves better accuracy in multiple evaluation metrics, including lower Mean Squared Error (MSE), higher Coefficient of Determination (R²), and improved Spearman’s Correlation (SPC), confirming its effectiveness in predicting drug responses and advancing personalized cancer treatment.</p>