Abstract Testing AI-Predicted Protein Motifs that Direct Constitutive Genomic AR Activity in Endocrine-Resistant Breast Cancer Ashfia F. Khan 1,2, Anthony S. Peidl 1,2, Shaymaa Bahnassy 3, Henry Vo2, Micah B. Castillo 2, Sarah K Herzog4,5, Suzanne AW Fuqua4,6, Preethi Gunaratne 2, Xiaolian Gao 2, Subash C. Pakhrin7, Tasneem Bawa-Khalfe1,2 1 Center for Nuclear Receptors & Cell Signaling, University of Houston, Houston, TX 2 Department of Biology & Biochemistry, University of Houston, Houston, TX 3 Department of Oncology, Lombardi Comprehensive Cancer Center, Georgetown University, Washington, DC 4 Lester & Sue Smith Breast Center, Baylor College of Medicine, Houston, TX 5 Program in Integrative Molecular and Biomedical Sciences, Baylor College of Medicine, Houston, TX 6 Dan L Duncan Comprehensive Cancer Center, Baylor College of Medicine, Houston, TX 7 Department of Computer Science & Engineering Technology, University of Houston-Downtown, Houston, TX Background: Endocrine therapy (ET) remains the first-line treatment for hormone-receptor positive (HR+) breast cancer (BCa). Approximately 15–20% of HR+ BCa are intrinsically resistant to ET, and 30–40% of patients acquire resistance. Resistance to ET (ET-R) supports cancer progression with reduced disease-free survival and greater incidence of metastatic disease. Hence, therapeutic strategies for ET-R HR+ BCa remain an overarching challenge. The androgen receptor (AR) is emerging as an attractive alternative target for BCa subtypes, and elevated AR levels drive HR+ BCa progression. Targeting AR in HR+ BCa is proving difficult with preclinical studies showing conflicting results for AR antagonists. Yet clinical trials with several AR-targeting drugs are ongoing. Our recent report highlights a unique constitutively active modified AR population that drives HR+ BCa metastatic properties and is insensitive to AR inhibitors. Our current objectives are to 1) use a novel machine-learning model to predict AR modifications and 2) establish a strategy to identify patients with high modified AR levels. Methods: An advanced artificial intelligence (AI) tool and mid-throughput microfluidic peptide array technology were used to map modification domains on AR. SUMO post-translational modification of AR (SUMO-AR) was eliminated in HR+ BCa using CRISPR-Cas9 technology. RNA-seq was employed to identify a unique gene signature for SUMO-AR, and comparative bioinformatic analysis stratified patients with high versus low SUMO-AR. Results: A novel deep-learning AI platform SumoPred-PLM is trained to identify consensus, non-consensus, and SUMO2/3-specific motifs on AR. We verified SUMO2/3-specific sites on AR with a mid-throughput microfluidic peptide array. The identified SUMO2/3-acceptor site of AR is important for HR+ BCa cell pathophysiology; loss of this SUMO2/3-acceptor site impacts endogenous AR SUMOylation, cell morphology, and proliferation/apoptosis. Using both high and low SUMO-AR BCa lines, a unique SUMO-AR gene profile was established. Our SUMO-AR gene signature identifies HR+ BCa patients with greater susceptibility to metastatic progression. Conclusion: Our studies present a unique pipeline that incorporates deep-learning AI technology to identify vulnerable motifs in AR for future drug discovery. Drug screens are currently ongoing. In addition, we establish a SUMO-AR gene signature that stratifies HR+ BCa patients with high/low SUMO-AR and predicts disease progression. We expect the results could be utilized to identify responders to AR inhibitors in ongoing clinical trials. Citation Format: Ashfia Khan, Anthony Peidl, Shaymaa Bahnassy, Henry Vo, Micah Castillo, Sarah Herzog, Suzanne Fuqua, Preethi Gunaratne, Xiaolian Gao, Subash Pakhrin, Tasneem Bawa-Khalfe. Testing AI-Predicted Protein Motifs that Direct Constitutive Genomic AR Activity in Endocrine Resistant Breast Cancer [abstract]. In: Proceedings of the 2023 San Antonio Breast Cancer Symposium; 2023 Dec 5-9; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2024;84(9 Suppl):Abstract nr PO5-05-09.