Abstract Introduction: Second-generation hormonal drugs have demonstrated significant efficacy in extending the survival of patients with advanced prostate cancer. These drugs operate through two distinct molecular mechanisms: inhibition of androgen biosynthesis and blockade of the androgen receptor. However, due to concerns of side effects, concurrent usage of multiple drugs is not recommended. A notable challenge is the eventual development of drug resistance, rendering all second-generation hormonal therapies ineffective. Currently, there is a lack of guidelines for helping patients choose the most suitable drug. In the context of pharmacogenomics, genetic differences can influence drug susceptibility and efficacy, with variations in genomic sequences and regulatory mechanisms impacting drug response at the transcriptional and translational levels. Considered the rapid processing capabilities of Affymetrix transcriptomic microarrays, we hypothesized that distinct gene signatures in patients who respond well or poorly to second-generation hormonal drugs could be identified and could act as a companion diagnostic tool. Methods: We enrolled 180 advanced prostate cancer patients, who fulfilled Taiwan national health insurance criteria of using second-generation hormonal drugs, from seven medical centers or regional hospitals in Taiwan at 3 time points: prior to drug administration, 3 months after treatment started and drug resistance developed. The RNA expression data and clinical parameters were analyzed, and classification models were built using IPA and R. Good or poor responders were defined by 3-month prostate-specific antigen (PSA) lowered by 90 percent. Results:: Differential expression and enrichment analyses post-treatment highlighted the activation of NF-kappa B signaling and various immune cell pathways. Notably, pathways related to primary immunodeficiency and IL-17 became prominent following the development of drug resistance. At the initial stages, distinct alterations in cell cycle pathways, particularly G1/S and G2/M, were observed between good and poor responders. This was elucidated through principal component analysis and subsequent dbscan clustering. To further distinguish between these responder groups, we utilized an integrative approach for feature selection, combining support vector machine, random forest classifier, and lasso regression. Despite the innovative approach, the model's error rate stood at approximately 15%, attributed mainly to the limited number of cases. Conclusion: Our study highlights the potential of using transcriptomic analysis to predict responses to second-generation hormonal drugs in advanced prostate cancer, paving the way for personalized treatment strategies. However, the need for larger cohort studies is evident to refine these predictive models and enhance their accuracy. Citation Format: Li-Hsin Chang, Hao-Han Chang, Xin-Yi Lin, Shu-Chi Wang, Chih-Pin Chuu, Deng-Neng Chen, Shu-Pin Huang, Chia-Yang Li. Personalized transcriptomic profiling for advanced prostate cancer: Guiding second-generation hormonal drug selection [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 5651.