Guo Chao, Li Zhen-Ling
Department of Hematology, China-Japan Friendship Hospital, Beijing, China.
Front Genet. 2025 Apr 4;16:1556519. doi: 10.3389/fgene.2025.1556519. eCollection 2025.
Identifying uncommon neutrophilic leukemias presents a challenging task, owing to the analogous morphological characteristics and the dearth of molecular markers. The transcriptomic profile of bone marrow cells in this disease subset has been rarely explored.
The OHSU-CNL dataset, encompassing clinical parameters and parallel transcriptomic matrix, was downloaded from the Genomic Data Commons (GDC) database. Distinctive co-expressed gene modules and pivotal genes for chronic neutrophilic leukemia (CNL) were identified using R software. Subsequently, a diagnostic model for CNL denoted as CNL-5 was formulated employing least absolute shrinkage and selection operator (LASSO) regression analysis. The diagnostic power of the CNL-5 model was compared with conventional clinical/genetic markers via multi-ROC analysis. The divergence in overall survival between CNL-5 risk groups was delineated by Kaplan-Meier analysis, and the predictive power (AUC and Harrison's C index) was determined by time-dependent ROC. Cell signaling pathways associated with CNL-5 risk were identified by genomic set enrichment analysis (GSEA).
Neither clinical indicators nor genetic markers were sufficient to classify neutrophilic leukemias. Through weighted gene co-expression network analysis (WGCNA), the brown module was discerned to be CNL-specific (p = 8e-16, R = 0.5). Using LASSO analysis, the CNL-5 model, with risk scores based on the weighted expression value of five genes (PDCD7/CR2/ZSCAN20/TRIM68/LILRA6) dichotomized patients into CNL-like and Atypical-CNL groups. Compared to the Atypical-CNL group, the CNL-like group demonstrated a clinical phenotype more consistent with CNL and had a significantly higher prevalence of CSF3R mutations (p < 0.05). Additionally, the AUC of the CNL-5 risk model surpassed that of conventional clinical/genetic markers, as validated by the GSE42731 dataset. Poorer survival was revealed in the high-risk group than in the low-risk group defined by the CNL-5 model. GSEA identified CNL-5-associated pathways, such as the inhibition of oxidative phosphorylation and the activation of IL6-JAK-STAT3 signaling.
A novel expression signature-based diagnostic assessment for CNL was developed, which showed better diagnostic utility than conventional indicators.
由于形态学特征相似且分子标志物匮乏，鉴别罕见的嗜中性粒细胞白血病是一项具有挑战性的任务。该疾病亚组中骨髓细胞的转录组特征鲜有研究。
从基因组数据共享库（GDC）数据库下载包含临床参数和平行转录组矩阵的OHSU-CNL数据集。使用R软件识别慢性嗜中性粒细胞白血病（CNL）独特的共表达基因模块和关键基因。随后，采用最小绝对收缩和选择算子（LASSO）回归分析构建了一个名为CNL-5的CNL诊断模型。通过多ROC分析将CNL-5模型的诊断能力与传统临床/基因标志物进行比较。通过Kaplan-Meier分析描绘CNL-5风险组之间总生存期的差异，并通过时间依赖ROC确定预测能力（AUC和哈里森C指数）。通过基因组集富集分析（GSEA）识别与CNL-5风险相关的细胞信号通路。
临床指标和基因标志物均不足以对嗜中性粒细胞白血病进行分类。通过加权基因共表达网络分析（WGCNA），发现棕色模块是CNL特异性的（p = 8e-16，R = 0.5）。使用LASSO分析，基于五个基因（PDCD7/CR2/ZSCAN20/TRIM68/LILRA6）的加权表达值计算风险评分的CNL-5模型将患者分为CNL样组和非典型CNL组。与非典型CNL组相比，CNL样组表现出与CNL更一致的临床表型，且CSF3R突变的患病率显著更高（p < 0.05）。此外，如GSE42731数据集所验证的，CNL-5风险模型的AUC超过了传统临床/基因标志物。CNL-5模型定义的高风险组的生存期比低风险组更差。GSEA识别出与CNL-5相关的通路，如氧化磷酸化的抑制和IL6-JAK-STAT3信号通路的激活。
开发了一种基于新型表达特征的CNL诊断评估方法，其诊断效用优于传统指标。
