e17026 Background: Androgen receptor (AR) signaling is a critical pathway in prostate cancer progression. Reactivation of the AR pathway is a key driver in mCRPC. Patients with mCRPC often develop resistance to novel AR signaling inhibitors (ARSIs) such as enzalutamide. AR inhibitor resistance leads to shorter overall and progression-free survival and can be attributed to AR gene amplification with subsequent AR overexpression and androgen-sensitization in prostate tumors. Identifying novel modulators of AR amplification in patients progressing on ARSIs may lead towards potential targets in drug discovery. Methods: In this cross-sectional study, an in silico patient model was developed on prescribed drug therapies and failed therapies patient information, PSA measures, ̃100 mutation variables including somatic, germline, and gain/loss mutations, and ̃1K gene expressions data from 275 mCRPC patients obtained from the Tempus database using GNS Healthcare’s Causal artificial intelligence technology, REFS. A causal network model was reconstructed based on patient data with inferred effect estimates. The network model was topologically surveyed for biological interactions. Results: Causal analysis of this model uncovered two germline mutations, located in PTCH1 and AXIN2, which modify the effect of AR copy number gain ( ARgain+ : copy number 6, ARgain- : copy number < 6) on AR gene expression. For PTCH1-mutant group, there was a 2-fold increase in AR gene expression in ARgain+ patients compared to ARgain- patients (p = 2.1e-12), while only 1.4 fold increase in AR gene expression from the PTCH1 non-mutant group (p = 6.3e-08). Similarly, there was a 2-fold increase in AR gene expression in ARgain+ patients compared to ARgain- patients (p = 7.2e-16) for the AXIN2-mutant group, while only one-fold increase in AR gene expression with the AXIN2 non-mutant group (p = 0.001). The C to T missense mutation in codon 3746 in PTCH1 and codon 148 in AXIN2 have higher prevalences, 55% and 72%, respectively in this mCRPC cohort compared to their population allele frequencies, 38% and 47%, respectively. Conclusions: PTCH1 and AXIN2 mutations from our analyses is consistent with a literature reported role for Hedgehog signaling and Wnt/-catenin signaling pathways, respectively, influencing the disease progression in castration-resistant prostate cancer. Our findings suggest that the missense germline mutations in PTCH1 and AXIN2 predicts response to enzalutamide and other ARSIs in CRPC setting, and propose potential targets for combination therapy to overcome ARSI drug resistance in hormone-refractory prostate cancer. [Table: see text]