Abstract Introduction The prostate exhibits a unique metabolism that changes during initial neoplasia to aggressive prostate cancer (PCa) and metastasis. The study of PCa metabolism thus represents a new avenue for diagnostics, particularly early diagnosis of aggressive PCa cases. Results Here, using transcriptomics data from The Cancer Genome Atlas (498 PCa patients), we identified six metabolic subgroups (C1-C6) of PCa that showed distinct disease-free survival outcomes (p<0.0001). In particular, we identified at least two PCa subgroups (C5 and C3) that exhibited significant poor prognosis (~70% and 30-40% relapse by the first 72 months; hazards ratios 9.4 and 4.4, respectively, relative to the best prognosis cluster C4 that showed <20% relapse even by 120 months). The subgroups were reproducible in an independent dataset from Taylors et al. 2010 (215 patients; p=0.00088). The subgroups displayed distinct metabolic profiles vis-à-vis normal tissues; measured as ‘deregulation’ of metabolic pathways (using Pathifier, Drier & Domany, 2013). In particular, the poor-prognosis subgroups C5 and C3 showed considerable deregulation for pathways involved in synthesis and catabolism of complex forms of lipids and carbohydrates, amino acids, and TCA cycle, and these were exhibited in parallel or in the face of glycolysis, a common form of energy production in cancer cells. Furthermore, the subgroups were significantly over-enriched for different sets of genetic alterations [particularly, deletions/mutations in BRCA1 and TP53 (C5), RB1 and STK11(C3); and AR amplifications (C1); p≤8.6E-04], suggesting that distinct alterations may be underpinning the subgroups and ‘pushing’ the subgroups towards their unique metabolic profiles. Finally, applying the classifier to blood expression profiles from 42 active surveillance (AS) and 65 advanced castrate resistant PCa (ACRPC) patients determined based on prostate-specific antigen (PSA) levels (Olmos et al. , 2012) assigned 70.77% ACPRC, and interestingly reassigned 59.52% AS patients to at least one of the poor prognosis subgroups (C5, C3) with 35.71% to the poor and metabolically deregulated subgroup C3. Conclusion The identification of PCa subgroups displaying distinct clinical outcomes solely from metabolic expression profiles of PCa tumours reiterates the significant link between deregulated metabolism and PCa outcomes (Eidelman et al. , 2017). On the other hand, the time to biochemical relapse (rise in PSA levels) was not indicative of the early relapse seen for the metabolically deregulated subgroups C3 and C5 (these show considerably late BCR compared to C4). Our study thus highlights specific processes (elevated lipid and carbohydrate metabolism pathways) that could be better indicators than PSA for early diagnosis of aggressive PCa.