标题： 基于深度学习的特征发现用于解码儿童高级别胶质瘤单细胞转录组学中的表型可塑性 显示英文标题

标题： Deep Learning-based Feature Discovery for Decoding Phenotypic Plasticity in Pediatric High-Grade Gliomas Single-Cell Transcriptomics

摘要： 通过使用复杂数学网络动力学和基于图的机器学习，我们确定了儿童高级别胶质瘤（pHGGs）亚型单细胞转录组中谱系特异性可塑性的关键决定因素：IDHWT胶质母细胞瘤和K27M突变胶质瘤。我们的研究识别了通过肿瘤-免疫微环境调控胶质瘤形态发生网络相互作用，包括神经发育程序、钙动态、铁代谢、代谢重编程以及MAPK/ERK和WNT信号之间的反馈回路。这些关系突显了在破坏的神经分化层次中导航的细胞状态混合光谱的出现。我们在IDHWT胶质母细胞瘤中识别出过渡基因，如DKK3、NOTCH2、GATAD1、GFAP和SEZ6L，在K27M亚型中识别出H3F3A、ANXA6、HES6/7、SIRT2、FXYD6、PTPRZ1、MEIS1、CXXC5和NDUFAB1。我们还识别出MTRNR2L1、GAPDH、IGF2、FKBP变体和FXYD7作为影响两个子系统细胞命运决定的过渡基因。我们的研究结果表明，pHGGs在表现出适应不良行为和混合细胞身份的状态中发育停滞。实际上，肿瘤异质性（亚稳态）和可塑性作为对免疫炎症微环境和氧化应激的应激反应模式出现。此外，我们展示了pHGGs主要由放射状胶质细胞的发育轨迹引导，优先选择新皮层细胞命运，在端脑和前额叶皮层（PFC）分化中。通过解决潜在的模式形成过程和可塑性网络作为治疗弱点，我们的研究结果提供了旨在调节胶质瘤细胞命运并克服治疗耐受性的精准医学策略。我们建议向神经元样谱系分化进行过渡治疗作为一种潜在的治疗方法，以帮助稳定pHGG的可塑性和侵略性。 显示英文摘要

摘要： By use of complex network dynamics and graph-based machine learning, we identified critical determinants of lineage-specific plasticity across the single-cell transcriptomics of pediatric high-grade glioma (pHGGs) subtypes: IDHWT glioblastoma and K27M-mutant glioma. Our study identified network interactions regulating glioma morphogenesis via the tumor-immune microenvironment, including neurodevelopmental programs, calcium dynamics, iron metabolism, metabolic reprogramming, and feedback loops between MAPK/ERK and WNT signaling. These relationships highlight the emergence of a hybrid spectrum of cellular states navigating a disrupted neuro-differentiation hierarchy. We identified transition genes such as DKK3, NOTCH2, GATAD1, GFAP, and SEZ6L in IDHWT glioblastoma, and H3F3A, ANXA6, HES6/7, SIRT2, FXYD6, PTPRZ1, MEIS1, CXXC5, and NDUFAB1 in K27M subtypes. We also identified MTRNR2L1, GAPDH, IGF2, FKBP variants, and FXYD7 as transition genes that influence cell fate decision-making across both subsystems. Our findings suggest pHGGs are developmentally trapped in states exhibiting maladaptive behaviors, and hybrid cellular identities. In effect, tumor heterogeneity (metastability) and plasticity emerge as stress-response patterns to immune-inflammatory microenvironments and oxidative stress. Furthermore, we show that pHGGs are steered by developmental trajectories from radial glia predominantly favoring neocortical cell fates, in telencephalon and prefrontal cortex (PFC) differentiation. By addressing underlying patterning processes and plasticity networks as therapeutic vulnerabilities, our findings provide precision medicine strategies aimed at modulating glioma cell fates and overcoming therapeutic resistance. We suggest transition therapy toward neuronal-like lineage differentiation as a potential therapy to help stabilize pHGG plasticity and aggressivity.