In a study published in Nature Communications, researchers from the Pan Prostate Cancer Group analysed whole-genome sequencing data from 961 primary prostate tumours, drawn from a cohort of 1,001 men across six countries: Canada, Germany, the United Kingdom, Australia, Denmark and France. Using standardised consensus pipelines, the team integrated six distinct classes of somatic alteration—copy-number gains and losses, protein-coding single-nucleotide variants and indels, non-coding small mutations, structural variant breakpoints, biallelic inactivation of tumour suppressor genes, and focal amplifications—to systematically detect recurrent genomic alterations, the statistical fingerprints of natural selection acting on cancer cells.

The scale of the analysis was formidable. The dataset encompassed more than 2.3 million small mutations, nearly 149,000 structural variants and over 160,000 copy-number events. Beyond protein-coding exons, the researchers examined roughly 16,000 non-coding RNA genes, untranslated regions, and more than 650,000 gene-regulatory and chromatin architectural elements mapped from prostate tumour epigenomics. This breadth mattered: nearly every tumour in the cohort—97 percent—carried at least one recurrent driver alteration, with an average of 8.4 per genome. Copy-number alterations dominated the landscape, appearing in 91 percent of tumours, followed by structural variants in 69 percent. By contrast, protein-coding driver mutations appeared in only about a third of tumours, underscoring why exome-based studies have repeatedly hit a ceiling in this disease.

The canonical landmarks of prostate cancer genomics were all confirmed at scale. Losses of NKX3-1 on chromosome 8p21 appeared in 58 percent of tumours, PTEN loss in 35 percent, TP53 loss in 33 percent and RB1 loss in 40 percent, while gains targeting the MYC oncogene at 8q24 occurred in 16 percent. ETS family oncogene fusions—chiefly ERG—were detected in roughly half of all tumours, consistent with earlier reports. But the integrative approach also surfaced 12 previously unreported driver candidates among the 53 recurrent alterations found in at least 1 percent of the cohort, and critically, 20 of the 53 would have been invisible to standard gene-panel or exome-sequencing methods, a pointed argument for whole-genome sequencing in clinical settings.

Among the most striking discoveries was a recurrent structural variant hotspot at the SLC30A4 gene on chromosome 15, altered in 26 tumour samples. SLC30A4 encodes a zinc transporter with known ties to prostate biology, but the structural variants appeared to do something unexpected: rather than disrupting the gene itself, they hijacked its enhancers. Most of these variants were translocations that physically relocated SLC30A4’s regulatory elements—marked by active chromatin signatures such as H3K27ac and BRD4 binding—next to distant oncogenes including ERG, ETV1, ETV4 and MYC. Genome-wide, the team found six additional solute carrier genes whose enhancers were similarly rearranged, affecting 2.8 percent of tumours overall. The pattern suggests an androgen-receptor-mediated mechanism in which structural variants co-opt tissue-specific enhancers early in tumourigenesis to switch on cancer genes, a process the authors describe as enhancer hijacking.

The study also quantified how driver profiles differ by ancestry and disease stage. Comparing 113 tumours from men of African ancestry with 927 non-African tumours, the researchers found African-ancestry cancers depleted of ETS alterations—13 percent versus 56 percent—while enriched for MYC gains, present in 40 percent versus 17 percent, along with gains of MDM4 and GATA2 and mutations in CDK12 and KMT2C. This shift toward a more aggressive alteration profile aligns with the disproportionately higher burden of lethal prostate cancer documented in men of African ancestry. Meanwhile, comparison with 103 metastatic tumours showed that biallelic PTEN inactivation appeared in half of all metastases, a 3.7-fold enrichment over primary disease, alongside AR amplifications and gains of MYC, MDM4 and COL5A1.

Pathway-level analysis revealed that these recurrent alterations converge on a limited set of biological programs. Hypoxia response genes were altered in 76 percent of the cohort, while PI3K/AKT, p53, androgen receptor and TGF-beta signalling were each disrupted in a third to two-thirds of tumours. Homologous recombination repair genes, including BRCA1, BRCA2, ATM and RAD51B, were frequently affected—clinically significant, because such alterations qualify patients for PARP inhibitor therapy. Indeed, when the team cross-referenced their findings against the OncoKB and CIViC databases, 23 percent of tumours carried at least one genome-informed therapeutic target with level 1 evidence, and 12 tumours showed tumour mutational burden high enough to suggest potential immunotherapy responsiveness.

Perhaps the most clinically consequential result emerged from unsupervised clustering of driver profiles, which resolved the cohort into four molecular subtypes the authors call oncotypes. Two carried unfavourable prognoses. Oncotype 4, termed TP53/LOH, was defined by widespread copy-number losses, chromosomal instability and frequent loss of PIK3R1, NKX3-1, TP53 and PTEN, with a 10-year metastasis incidence of 35 percent. Oncotype 2, termed MYC/BRCA2, combined chromosome 8q and MYC gains with biallelic inactivation of CHD1 and BRCA2 and mutations in CDK12 and SPOP, showing a 22 percent ten-year metastasis rate and notable enrichment among donors of African ancestry. At the other extreme, the largest group—oncotype 1, the RGA-desert—carried few alterations and only a 5 percent ten-year metastasis incidence, while oncotype 3 was dominated by ETS fusions and PTEN loss at intermediate risk.

Validation strengthened the oncotype framework. Mapping the same recurrent alterations onto an independent cohort from The Cancer Genome Atlas, the researchers trained a random forest classifier that reproduced the four subtypes and confirmed their association with progression-free intervals. Within the discovery cohort, the oncotypes separated significantly by both metastasis-free survival and biochemical recurrence, even after adjustment for Gleason grade, tumour stage and age at diagnosis. Individual drivers also carried prognostic weight: protein-coding TP53 mutations, MDM4 and COL5A1 gains, FOXA1 structural variants and BRCA1 loss each independently predicted shorter metastasis-free survival in multivariable Cox models, with hazard ratios ranging from roughly 3 to nearly 6.

The authors are careful about limitations. Sequencing depth and intratumoural heterogeneity likely mean the true driver landscape is underestimated, and the cohort is predominantly of European ancestry, leaving the genomic picture in other populations incomplete. Still, the message is clear: prostate cancer’s driver repertoire extends far beyond the protein-coding exome, into structural rearrangements, enhancer hijacking and non-coding regulatory elements, and these patterns carry real prognostic and therapeutic weight. For a disease in which overtreatment of indolent tumours remains a central clinical dilemma, a whole-genome taxonomy that separates the RGA-desert from the TP53/LOH and MYC/BRCA2 oncotypes offers a concrete path toward more precise risk stratification—and, potentially, toward targeted interventions for the patients who need them most.

Subject of Research: Genomic driver alterations and molecular heterogeneity in primary prostate cancer

Article Title: The molecular heterogeneity and clinical impact of genomic driver alterations in prostate cancer

Article References: Pellegrina, D., Kiriy, D., C. L. Cheng, K., Favero, F., Miletti, E., Klein, Z. P., Hernando, B., Olsen, A. V., Sahli, A., Fu, Y., Lutsik, P., Jaratlerdsiri, W., Hutchison, W. J., Bova, G. S., Brewer, D. S., Brook, M., Brors, B., Butler, A., Cancel-Tassin, G., … Reimand, J. (2026). The molecular heterogeneity and clinical impact of genomic driver alterations in prostate cancer. Nature Communications, 17(1), Article 10572. https://doi.org/10.1038/s41467-026-77900-7

Image Credits: AI Generated

DOI: 10.1038/s41467-026-77900-7

Keywords: prostate cancer, cancer genomics, whole-genome sequencing, structural variants, enhancer hijacking, oncotypes, tumour suppressor genes, copy-number alterations, metastasis-free survival, precision oncology, SLC30A4, ancestry differences