Artificial intelligence and multi-omics integration in liquid biopsy for genitourinary cancers: a systematic scoping review.

BACKGROUND: Liquid biopsy, combined with multi-omics profiling and artificial intelligence (AI), offers a minimally invasive strategy for cancer detection, monitoring, and prognostication. Evidence in genitourinary (GU) cancers remains fragmented across analytes, biofluids, and computational approaches, limiting clinical translation. OBJECTIVE: To map and synthesize AI-enabled liquid biopsy research in GU cancers and identify methodological and clinical gaps. METHODS: A systematic scoping review was conducted following PRISMA-ScR and JBI guidelines (protocol registered in PROSPERO: CRD420251137304). Comprehensive searches of MEDLINE, Embase, Scopus, Web of Science, Cochrane Library, IEEE Xplore, preprint servers, and major conference proceedings identified eligible studies published 1 January 2018-1 August 2025. Dual independent screening and data charting were performed. Note on performance metrics: reported AUC, sensitivity, and specificity values were extracted verbatim from included studies and from previously published meta-analyses when cited; this review did not perform a de-novo quantitative meta-analysis or statistical pooling of primary study results. RESULTS: A total of 115 studies were included, spanning biomarker discovery, algorithm development, and early translational efforts. Urinary exosome diagnostics in urothelial cancer have been reported in prior meta-analyses to show pooled AUC ≈ 0.83 (sensitivity ≈ 75%, specificity ≈ 77%); these pooled estimates were cited but not recalculated in the present scoping review. ctDNA-based approaches showed strong prognostic value in advanced prostate cancer, with detection rates consistently > 60%. Testicular germ cell tumor research remains dominated by miR-371a-3p studies, with limited AI or multi-omics integration. Intermediate-fusion and graph-based models offered advantages over simple concatenation but were hindered by small sample sizes, batch effects, and lack of external validation. CONCLUSION: AI-driven multi-omics liquid biopsy holds promise for GU oncology but is constrained by fragmented evidence and methodological variability. Future research should prioritize standardized pipelines, multi-center validation, prospective trials, and explainable AI to accelerate clinical translation.