AI and Cancer Research ── Cancer biology ── 2026-11-08
Cancer molecular biology and AI
- Abstract
BACKGROUNDEsophagogastroduodenoscopy is operator-dependent, and early upper gastrointestinal neoplasms are frequently missed.Although randomized trials and a previous metaanalysis have reported that real-time artificial intelligence (AI) improves neoplasm detection, the first large multicenter trial with centralized pathology adjudication found no significant benefit for gastric neoplasms.Furthermore, the pooled evidence has not been reassessed following publication of this trial. AIMTo reappraise whether AI-assisted endoscopy increases per-patient neoplasm detection overall and by anatomical site. METHODSFollowing a registered protocol (PROSPERO CRD420261424675), we searched PubMed, EMBASE, and Cochrane Library through June 2026 and two Chinese-language databases through September 2026 for randomized trials comparing conventional and real-time AI-assisted endoscopy in adults.Risk of bias (RoB) and certainty of evidence were assessed using the RoB 2 tool and GRADE rating.Effects were pooled in Python using restricted maximum-likelihood random-effects models with Hartung-Knapp-Sidik-Jonkman confidence intervals (CIs).Prediction intervals were calculated, with sensitivity analyses by anatomical target and system functionality. RESULTSFourteen randomized trials (90828 participants) were included.AI increased per-patient detection of upper gastrointestinal neoplasms [10 trials; risk ratio (RR): 1.47, 95%CI:1.25-1.73;I 2 = 40%; prediction interval 1.04-2.09;moderate certainty], with similar 4/51 esophageal (RR: 1.61, 95%CI: 1.27-2.03;I 2 = 0%; moderate certainty) and gastric (RR:1.45, 95%CI: 1.11-1.89;I 2 = 60%; low certainty; subgroup P = 0.50) estimates and no variation by system functionality.Patient-level premalignant lesion detection increased (4 trials; RR: 1.80, 95%CI: 1.09-2.96),early cancer detection was imprecise (6 trials; RR:1.28, 95%CI: 0.86-1.91),and the lesion-level miss rate was lower with AI (RR: 0.21, 95%CI: 0.09-0.49).Adverse events did not differ; heterogeneous procedural outcomes were synthesized narratively.The largest, centrally adjudicated trial drove the gastric heterogeneity. CONCLUSION AI-assisted endoscopy increases upper gastrointestinal neoplasm detection in Asiantrial settings, although the evidence for gastric neoplasia remains heterogeneous and of low certainty.Further confirmatory trials with patient-important endpoints are warranted.
Journal IF-equivalent: 4.7 (OpenAlex 2-year mean citedness, value as of 2026-11-06, retrieved 2026-11-07; not the official Clarivate IF)Reference: Choi Sj, Chae G. Artificial intelligence-assisted upper gastrointestinal endoscopy for neoplasm detection: Systematic review and meta-analysis of randomized controlled trials. World J Gastroenterol. 2026 Nov 4 [Epub ahead of print]. doi:10.3748/wjg.127326.Checked: Abstract only - Abstract
BACKGROUND: Estimating tumor fraction from whole-genome cell-free DNA sequencing is critical for liquid biopsy, but is hampered by weak signals and baseline noise at low tumor fractions. Existing computational methods often require matched controls or large labeled datasets for training and lack uncertainty quantification. To address these gaps, we developed purNPE, a Bayesian deep-learning framework trained without labeled cancer cell-free DNA samples. Specifically, purNPE leverages a two-part generative model: one component simulates diverse tumor copy-number profiles based on evolutionary genealogies, while a second, data-driven component learns and replicates realistic sequencing background patterns from cancer-free cell-free DNA. By training a Neural Posterior Estimator on synthetic tumor profiles augmented with learned noise, purNPE performs amortized inference in milliseconds without needing a reference sample set at inference. RESULTS: In a real-world pan-cancer cohort, purNPE achieved comparable performance with existing methods against an orthogonal mutant-allele-fraction proxy (MAE [Formula: see text]). In semi-synthetic and in silico experiments, purNPE showed calibrated uncertainty estimates and a dose-response across low tumor-fraction spike-ins, with separation of approximately 1% TF spike-ins from blanks under the evaluated conditions. CONCLUSIONS: This work provides a proof-of-concept framework for using simulation-based inference to derive calibrated, uncertainty-aware TF estimates. Larger prospective cohorts with stronger orthogonal ground truth are required before analytical sensitivity or clinical utility can be established.
Journal IF-equivalent: 7.7 (OpenAlex 2-year mean citedness, value as of 2026-11-06, retrieved 2026-11-07; not the official Clarivate IF)Reference: Volkov H, Raitses-Gurevich M, Grad M, Shlayem R, Danilevsky A, Rubinek T, et al. Simulation-based Bayesian deep learning enables uncertainty-aware tumor fraction estimation in cell-free DNA. BioData Mining. 2026 Aug 6;19(1):84. doi:10.1186/s13040-026-00595-5.Checked: Abstract only