*3.7. WGM Derived Data Filtering*

Filtering steps were performed on the resulting SVs. First, SVs that did not pass the Bionano recommended confidence level for the corresponding SV type were excluded; that is, all SVs other than inversion must have confidence > 0.1 and both breakpoints of an inversion event must have confidence > 0.01. Second, only rare SVs were included, defined as being observed in < 10% of a cohort of ~150 "normal" samples provided by Bionano [121]. Finally, "somatic" SVs were identified as those supported by a minimum of *yt* molecules in the tumour sample but not observed in more than *yn* molecules in the matching-normal sample, where *y* = −*0.3* + *0.13 \* x* and *x* being the effective coverage of the corresponding sample. This formula is recommended by Bionano as detailed in their Variant Annotation Pipeline v1.0 (BNG document # 30190). The minimum coverage cut-offs for somatic SV calling are summarized in Table S8. The WGM data are available at the following Doi: 10.25833/7wqs-gb12 [122].

#### *3.8. Generation of a Prostate Cancer-Related Gene List*

In addition to identifying annotated pathogenic and likely pathogenic alterations as well as the top genes affected by SCNAs, we reviewed alterations involving potential PCa driver genes and non-coding events associated with prostate cancer. A list of 159 PCa-associated genes was compiled from recent studies that identified recurrently mutated genes in primary and metastatic samples (Table S7) [6–9]. The list included commonly altered genes with potential functional relevance from The Cancer Genome Atlas (TCGA) primary PCa data [7,123], potential driver genes identified in primary and metastatic samples by Wedge et al. [8] and genes recurrently mutated in metastatic disease as identified by Robinson et al. [6] and Armenia et al. [9]. A list of non-coding events was compiled from recent published data (Table 4).

#### *3.9. Other Analyses*

The full list of binding clusters of 340 TFs compiled by the ENCODE project was obtained from the University of California Santa Cruz data repository (encRegTfbsClistered table; last updated 16 May 2019) and examined for somatic variants using a custom R script. Somatic variants within AR binding sites were evaluated against published putative binding sites observed in the LNCaP prostate cancer cell line (NCBI Gene Expression Omnibus accession GSE83860; [33]). The Circos plots in Figures 3–7 were generated using the CIRCOS software (v0.69-6) [124] based on SNV/indel data from MUTECT, copy number data from CNVkit, and SV data from GRIDSS. Phylogenetic reconstruction of tumour evolution for patient 19651 was performed using phyloWGS [116] based on SNV/indel data from MUTECT and copy number data from TITAN. Analyses of COSMIC Mutational Signature [125] clonal evolution was performed using the R package Palimpsest v1 [126] which utilized SNV data from MUTECT for estimates of mutation signature and SNV allele frequency data from MUTECT along with copy number segmentation data from Sequenza [106] for estimates of clonality.

#### **4. Discussion**

These real-world clinical cases demonstrate that clinically relevant mutations occur even in treatment-naïve patients across the spectrum of disease from high-risk primaries to metastatic cases. While the pathways impacted in these cases align with those identified in larger scale genomic studies, the coexistence of multiple alterations has not been explored. These findings raise several points.

Firstly, studies of neoadjuvant or adjuvant docetaxel in men with high-risk localized disease undertaking radical prostatectomy or definitive radiotherapy with ADT have had conflicting results [4, 46,71] but a subgroup of men appear to benefit. Poor prognostic genomic findings, such as *TP53* deletions or deleterious *BRCA2* alterations at baseline may be useful in selecting men for additional treatment. Similarly, not all men require escalated treatment beyond ADT for HSPC but biomarkers to guide treatment selection remain limited. The findings of *TP53* and/or Wnt pathway activating alterations in 5/8 (63%) of our primary samples highlight a potential biomarker for selecting men that should be considered for escalated therapy, preferably with docetaxel rather than a novel ASI [32,62,127]. Though speculative in the hormone sensitive setting, there is mounting evidence these alterations could be useful in guiding treatment selection in CRPC. Secondly, we observed events in minimally treated patients, such as *NCOR1* and *NCOR2* losses, that may be associated with ADT resistance. These alterations again may identify patients at risk for early development of CRPC who may need escalated therapy upfront. Thirdly, pathway mutations typically enriched in metastatic CRPC, particularly PI3K and MAPK pathway SCNAs, were frequently seen in our patients and represent potential targets for neoadjuvant intervention in high risk localized and/or de novo metastatic HSPC clinical trials.

The addition of WGM in our study did not identify a current therapeutic target but it did identify SVs impacting oncogenic and tumour suppressing genes that were not identified by using WGS alone. Though we did identify non-coding events affecting the promoters, enhancers and TF binding sites of relevant genes, their therapeutic relevance has yet to be elucidated. However, as WGS and WGM data accumulate and annotations improve, we may find new relevant mutations and begin to understand how they may be integrated into clinical practice. Additionally, the use of complementary genomic technologies such as RNA-sequencing and chromatin immunoprecipitation sequencing may improve our ability to translate genomic data into real-world clinical decision-making.

In this retrospective study, we assess the current status of genome profiling, specifically WGS and WGM, to inform decision-making for 13 patients presenting with metastatic PCa. Our findings suggest that, despite being a cancer associated with a low TMB, individual PCas can harbour complex series of mutations affecting multiple growth pathways. Therefore, the precision medicine model of identifying one target to treat is unlikely to succeed. Given its heterogeneity and despite comprising only a very small fraction of the I-PREDICT study cohort [128], PCa may be the ideal cancer to test the paradigm of using genomics to identify and treat multiple targets simultaneously.

#### **5. Conclusions**

Our analyses demonstrate that whole genomic interrogation of PCas may provide invaluable information at any stage of the disease. Most of our cases had alterations affecting multiple signalling pathways highlighting the utility of a comprehensive molecular assessment in tailoring treatment strategies to an individual. Moreover, WGM identified SVs disrupting prostate cancer relevant genes that were not apparent on our WGS analyses. Many non-coding and WGM events were identified but their therapeutic relevance is yet to be established. Though these data add to our current knowledge, further research is needed, potentially integrating additional genomic technologies, to identify new treatment targets and predictive biomarkers. While several potential biomarkers that may influence treatment decisions were found in these patients, most have yet to be validated in prospective clinical trials.

*Cancers* **2020**, *12*, 1178

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2072-6694/12/5/1178/s1: Table S1: Potentially deleterious germline SNVs annotated by ANNOVAR, Table S2: Sequencing and mapping data including TMB and PGA, Table S3: SNVs affecting selected driver genes as annotated by ANNOVAR, Table S4: SCNAs identified by CNVKit affecting selected driver genes, Table S5: SVs affecting selected driver genes, Table S6: SVs identified by WGM, Table S7: Selected driver genes, Table S8: WGM SV calling cut-offs. File S1: Custom WGM SV calling parameters.

**Author Contributions:** Conceptualization, V.M.H., E.K.F.C., M.C., A.M.J., C.A.M.J., A.A.K, C.M.H. and P.I.C..; methodology, V.M.H., M.C., E.K.F.C., W.J., T.G., R.J.L., and D.C.P.; formal analysis, V.M.H., M.C, E.K.F.C., W.J., T.G.; resources, V.M.H., M.C., C.A.M.J., P.D.S., N.C., R.J.L., A.M.F.K., C.A.M.J., A.-M.H., C.M.H., A.A.K.; data curation, M.C., E.K.F.C.; writing—original draft preparation, M.C.; writing—review and editing, V.M.H., M.C., E.K.F.C., A.M.J., A.M.F.K., W.J., T.G., C.M.H., R.J.L.; visualization, E.K.F.C., M.C., V.M.H.; supervision, V.M.H.; project administration, V.M.H.; funding acquisition, V.M.H., A.M.J, C.A.M.J., P.D.S., P.I.C., N.C., C.M.H. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was funded by the Movember Australia and the Prostate Cancer Foundation Australia (PCFA) as part of the Movember Revolutionary Team Award (MRTA) to the Garvan Institute of Medical Research on prostate cancer bone metastasis (ProMis led by P.I.C. with team leads V.M.H., N.C. and C.M.H.), and an Ian Potter Foundation infrastructure award to V.M.H.. M.C. and T.G. was funded by an Australian Government Research Training Program Scholarship, E.K.F.C. and D.C. by MRTA-ProMis, W.J. by the Australian Prostate Cancer Research Centre NSW (APCRC-NSW), and V.M.H. by the University of Sydney Foundation and Petre Foundation, Australia.

**Acknowledgments:** This work is supported by the HPC resources generously provided by the National Computational Infrastructure (Raijin/Gadi), the Garvan Institute of Medical Research (Wolfpack) and the University of Sydney (Artemis). We would like to acknowledge the Bionano team for generation of Case 19651's WGM data and additional support. We also thank the patients for their participation and the Garvan Institute Biobank and its staff for access to the tumor samples.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.
