Rédiger un PREreview

Background: Neuroblastoma (NBL) prognosis relies on clinical-molecular features that leave the 1-2 year age window—a challenging gray zone—poorly resolved. We developed a DNA-only trinucleotide mutation spectrum (TMS) signature to address this gap.Procedure: We trained TMS using lasso-penalized Cox regression on the OpenPedCan NBL MAF cohort (n=781), validated in TARGET (n=220) and GMKF (n=93) within the 1-2 year window, and compared against MYCN, mutation count, and RNA quality.Results: TMS captures a C>A/G>T-dominated oxidative damage spectrum, with ATC>C>A selected in 100/100 stability iterations. TMS_z correlated with oxidative repair genes, antioxidant genes, and C>A mutation burden (Spearman r=0.744). The prognostic effect was age-dependent, peaking at 1-2 years (1yr: HR=1.70, P=3.3e-5; 2yr: HR=1.47, P=9.0e-7; interaction P=3.4e-6). A threshold, TMS_z > 0.5, stratified patients (HR=3.13, 95% CI 2.10-4.66, P=1.9e-8; OR=5.09, P=1.2e-7), validated in TARGET (HR=2.96, P=4.0e-7) and directionally consistent in GMKF (HR=2.03, P=0.37). TMS remained significant after adjusting for MYCN (HR=3.65, P=0.0093) and mutation count (HR=1.73, P=2.2e-9), though partially overlapping with RNA quality axes (per-SD HR=1.32, P=0.063).Conclusions: TMS is a simple, DNA-only prognostic tool for 1-2 year NBL, providing MYCN- and mutation-count-independent risk stratification. Its DNA-only nature makes it applicable to FFPE and resource-limited settings.