5067 Background: Response to secondary hormonal manipulations and cytotoxic therapy is variable in men with castration resistant prostate cancer (CRPC). Androgen receptor (AR) signaling persists in CRPC but there are no established biomarkers which facilitate the choice between continued hormonal manipulations or initiation of cytotoxic therapy in CRPC. Methods: A transcription-based AR activity signature was developed using established analytic methods from an androgen-sensitive prostate cancer cell line (LNCaP). The AR signature assigns a probability of AR activity to individual samples (range 0 to 1) and was applied to independent datasets of LNCaP cells and human prostate samples with known androgen status. The AR signature was subsequently applied to multiple human prostate cancer datasets to determine if AR activity in individual tumor specimens correlates with clinical and molecular phenotypes. Results: A robust expression-based signature of AR activity can accurately predict the AR activity status of independent LNCaP samples, anticipates the hormonal status of individuals (castrate v non-castrate), and assigns an AR activity score that correlates with intraprostatic androgen levels in human biopsy specimens (two-tail Spearman p=0.022 for dihydrotestosterone). The mean difference between AR activity between replicates was minimal (5%) and intra-individual consistency across different metastatic sites was also consistent. In two independent datasets, AR activity was predicted to be higher in local, untreated prostate tumors (mean 0.80 ± 0.23, n=16 and 0.72 ± 0.21, n=10) and decreased but more heterogeneous in CRPC (0.42 ± 0.33, n=32 and 0.40 ± 0.33, n=10)(Mann Whitney P=0.0001 and 0.036, respectively). Low probability of AR activity correlated strongly with predicted SRC activity in local (p = 0.006, n=79) and metastatic (p=0.059, n=32) prostate tumors and predicted sensitivity to dasatinib, a src-targeting kinase inhibitor. Conclusions: Determining the level of AR activity in an individual’s CRPC has the potential to identify patients most likely to benefit from continued hormonal manipulation versus those who are more likely to benefit from other biologically targeted or cytotoxic therapies. No significant financial relationships to disclose.