Treatment decisions in metastatic castration-resistant prostate cancer (mCRPC) are mostly guided by clinical variables, but efforts to molecularly monitor the disease remain hampered by challenges in acquiring tumor tissue repeatedly. Here, we simultaneously profiled the genome copy number and exome in longitudinal plasma circulating tumor DNA (ctDNA) acquired before, during, and upon progression to serial treatments with androgen signaling inhibitors (ASI) and taxane chemotherapy from 60 mCRPC patients (2-10 samples per patient). The genomic data was used to delineate the clonal substructure and evolutionary dynamics of each patient, and an evolutionary dynamic index (EDI) was developed to measure the longitudinal changes of the tumor subclones. Treatment with ASI resulted in greater subclonal selection and population structure changes than taxane treatment. The subclones that emerged in association with serial therapy resistance harbored recurrent aberrations in previously identified and new candidate genes, with particular enrichment in genes related to PI3K-AKT signaling. These findings indicate that the integration of detailed clinical and genomics data can provide a framework for future unbiased genomic applications for ctDNA in the clinic to enable precision medicine.