Profiling of circulating tumor DNA (ctDNA) has the potential to provide key diagnostic and prognostic information to improve management of metastatic cancer. Blood plasma biopsies from metastatic cancer often contain ctDNA at high abundance, making high-throughput sequencing feasible from small blood volumes. Sequencing ctDNA can identify both point mutation and structural variants affecting key cancer genes.

We performed targetted high-throughput sequencing of ctDNA on blood plasma samples from patients with castrate resistant prostate cancer. Our analysis reveals reccurrent mutations in known mutated prostate cancer driver genes, including TP53, MLL2, FOXA1 and APC. In addition to analysing point mutations, we studied structural variation, as prostate cancer has been shown to frequently involve genomic rearrangments. To detect key structural variants with high confidence, we developed a novel bioinformatic algorithm for identifying and visualising these genomic events. Our algorithm combines orthogonal discordant read pair and split read evidence to call individual structural variants, and produces clear visualisation of individual events, assisting discrimination between true positives and false positives. When applied to paired-end sequence data from CRPC ctDNA blood plasma, our algorithm identifies high confidence structural variants. In particular, we identify large genomic deletions, inversions, tandem duplications and translations, at genomic locations of clinical significance.