背景：前列腺癌（PCa）高度异质性，使得精确区分肿瘤病变的临床阶段和组织学分级变得困难，从而导致过度和欠度治疗的大量发生。因此，我们期望开发新的预测方法以预防不适当疗法的发生。新兴证据显示溶酶体相关机制在PCa的预后中起着关键作用。本研究旨在鉴定PCa溶酶体相关的预后预测因子，以供未来治疗之用。 方法：本研究涉及的PCa样本来自癌症基因组图谱数据库（TCGA）（n=552）和cBioPortal数据库（n=82）。在筛选过程中，我们根据ssGSEA中位数分数将PCa患者分为两个免疫组。然后，采用单变量Cox回归分析和最小绝对收缩和选择操作（LASSO）分析，包括Gleason分数和溶酶体相关基因进行筛选。通过使用未经调整的Kaplan-Meier估计曲线和多变量Cox回归分析对进展自由间隔（PFI）的概率进行建模。使用受试者工作特征曲线（ROC）曲线、核算图和校准曲线来检验该模型在区分进展事件和非事件方面的预测价值。通过从队列中创建训练集（n=400）、内部验证集（n=100）和外部验证集（n=82）来反复验证该模型。 结果：在按ssGSEA分数分组后，Gleason分数和两个LRG——中性粒细胞细胞质因子1（NCF1）和γ-干扰素诱导的溶酶体硫醇还原酶（IFI30）被筛选出来，以区分具有或不具有进展风险的患者（1年AUC=0.787；3年AUC=0.798；5年AUC=0.772；10年AUC=0.832）。具有较高风险的患者表现出较差的预后结果（P<0.0001），并且具有较高的累积危险（P<0.0001）。此外，我们的风险模型将LRGs与Gleason分数相结合，与仅使用Gleason分数相比可以更准确地预测PCa预后。在三个验证集中，我们的模型仍然具有高的预测率。 结论：综上所述，这种新型的与溶酶体相关的基因签名与Gleason分数相结合，在PCa的预后预测中表现良好。

Background: Prostate cancer (PCa) is highly heterogeneous, which makes it difficult to precisely distinguish the clinical stages and histological grades of tumor lesions, thereby leading to large amounts of under- and over-treatment. Thus, we expect the development of novel prediction approaches for the prevention of inadequate therapies. The emerging evidence demonstrates the pivotal role of lysosome-related mechanisms in the prognosis of PCa. In this study, we aimed to identify a lysosome-related prognostic predictor in PCa for future therapies. Methods: The PCa samples involved in this study were gathered from The Cancer Genome Atlas database (TCGA) (n = 552) and cBioPortal database (n = 82). During screening, we categorized PCa patients into two immune groups based on median ssGSEA scores. Then, the Gleason score and lysosome-related genes were included and screened out by using a univariate Cox regression analysis and the least absolute shrinkage and selection operation (LASSO) analysis. Following further analysis, the probability of progression free interval (PFI) was modeled by using unadjusted Kaplan-Meier estimation curves and a multivariable Cox regression analysis. A receiver operating characteristic (ROC) curve, nomogram and calibration curve were used to examine the predictive value of this model in discriminating progression events from non-events. The model was trained and repeatedly validated by creating a training set (n = 400), an internal validation set (n = 100) and an external validation (n = 82) from the cohort. Results: Following grouping by ssGSEA score, the Gleason score and two LRGs-neutrophil cytosolic factor 1 (NCF1) and gamma-interferon-inducible lysosomal thiol reductase (IFI30)-were screened out to differentiate patients with or without progression (1-year AUC = 0.787; 3-year AUC = 0.798; 5-year AUC = 0.772; 10-year AUC = 0.832). Patients with a higher risk showed poorer outcomes (p < 0.0001) and a higher cumulative hazard (p < 0.0001). Besides this, our risk model combined LRGs with the Gleason score and presented a more accurate prediction of PCa prognosis than the Gleason score alone. In three validation sets, our model still achieved high prediction rates. Conclusion: In conclusion, this novel lysosome-related gene signature, coupled with the Gleason score, works well in PCa for prognosis prediction.Copyright © 2023 Huang, Yang, Zhang, Zhou, Duan, Liu, Li and Zhao.