Background: The optimal sequence to administer approved life prolonging ARS inhibitors: Abiraterone (A) and Enzalutamide (E),and taxanes: Docetaxel (D) and Cabazitaxel (C), to maximize individual pt benefit is an unmet medical need. Sensitivity to an individual agent is dependent on the intrinsic biology of the tumor and the treatment to which it has been exposed. For each agent, response rates are lower in pts who have received prior treatment with any of the other drugs, but for the individual, sensitivity to one does not necessarily predict sensitivity to the other. We sought to develop response and resistance signatures to A & E or D & C from a baseline blood sample. Methods: 91 pt blood samples collected from 79 pts for CTC analysis with the Epic Sciences platform prior to treatment (27 pre-A, 28 pre-E, 28 pre-D, 8 pre-C). Epic analysis identified traditional CTCs (CK+, CD45-, intact nuclei, morph distinct), CK- CTCs (CK-, CD45-, intact nuclei, morph distinct), small CTCs (CK+, CD45-, intact nuclei, small cell size), and CTC clusters. IF staining for AR N, AR C exp were performed, digital pathology algorithms analyzed CTC morphology. A classifier was developed to associate clinical phenotypes with outcome to a specific agent. Results: A and E signatures included: AR N/C exp. and presence of CK+, AR+, Nucleoli+ CTCs. D and C signatures by: presence of CK+, small, AR-, Nucleoli+ CTCs. Multivariate algorithms for A and E and D & C were statistically associated with de novo resistance. Line of therapy was not a univariate predictor of response. Conclusions: Characterization of CTCs identified predictive biomarkers of sensitivity to ARS Tx and taxane chemotherapy in mCRPC pts. The proposed A&E signature differs from C and D, providing the opportunity to better guide treatment selection and improve patient outcomes using a real-time, non-invasive blood biopsy. Prospective trials to validate results are planned.