PO.BCS01.01 · 生物信息与计算
人类各类癌症中m6A RNA甲基化的全面表征
Comprehensive characterization of m 6 A RNA methylation across human cancers
作者与单位 Authors & Affiliations
摘要 Abstract
中文摘要
N6-甲基腺苷（m6A）是最丰富的内部mRNA修饰，在基因调控、肿瘤进展和治疗应答中发挥着关键作用。然而，m6A RNA甲基化在人类癌症中的生物医学和临床意义仍未被完全理解。在此，我们使用m6A-seq分析，生成了一个全面的全转录组m6A修饰图谱，涵盖癌症基因组图谱（TCGA）中23种癌症类型、226个肿瘤样本的15,812个位点。为阐明m6A变异的调控决定因素和下游后果，我们将这些数据与广泛的多组学特征相整合。我们发现全局m6A模式分离为五个主要簇，主要由癌症谱系驱动。体细胞突变通过改变DRACH基序对局部m6A水平产生广泛而多样的影响。约10%的蛋白编码基因显示出m6A丰度与mRNA或蛋白表达之间一致的正向或负向关联。这些基因富集于转录因子，且其m6A水平强烈影响关键的肿瘤细胞状态，如上皮-间质转化（EMT）和缺氧。我们进一步表征了7482个TCGA样本中15个m6A调控因子的蛋白表达图谱，并揭示了由多种不同遗传和表观遗传机制引起的跨癌症频繁失调。最后，我们开发了一个深度学习模型，整合局部DNA序列背景、基因水平特征、m6A调控因子状态和肿瘤特异性背景，以预测m6A强度。该模型对相当一部分m6A位点实现了高准确性，能够对m6A变异进行大规模推断，并促进在广泛患者队列中的生物标志物发现。总之，这项研究为理解癌症中m6A甲基化的基因组图景和调控结构提供了关键资源，并为将m6A作为一类新的生物标志物和治疗靶点加以利用奠定了基础。
查看英文原文 English abstract
N6-methyladenosine (m 6 A) is the most abundant internal mRNA modification and plays essential roles in gene regulation, tumor progression, and therapeutic response. However, the biomedical and clinical significance of m 6 A RNA methylation in human cancer remains incompletely understood. Here, using m 6 A-seq profiling, we generated a comprehensive transcriptome-wide atlas of m 6 A modifications across 15,812 sites from 226 tumor samples spanning 23 cancer types in The Cancer Genome Atlas (TCGA). To elucidate the regulatory determinants and downstream consequences of m 6 A variation, we integrated these data with a broad spectrum of multi-omics features. We found that global m 6 A patterns segregate into five major clusters largely driven by cancer lineages. Somatic mutations exert widespread yet diverse effects on local m 6 A levels through alteration of DRACH motifs. Approximately 10% of protein-coding genes showed consistent positive or negative associations between m 6 A abundance and mRNA or protein expression. These genes are enriched for transcription factors, and their m 6 A levels strongly influence key tumor cell states such as epithelial-mesenchymal transition (EMT) and hypoxia. We further characterized the protein expression landscape of 15 m 6 A regulators in 7482 TCGA samples and uncovered frequent dysregulation across cancers arising from multiple distinct genetic and epigenetic mechanisms. Finally, we developed a deep-learning model that integrates local DNA sequence context, gene-level features, m 6 A regulator states, and tumor-specific context to predict m 6 A intensity. The model achieved high accuracy for a substantial fraction of m 6 A sites, enabling large-scale inference of m 6 A variation and facilitating biomarker discovery in extensive patient cohorts. Together, this study provides a key resource for understanding the genomic landscape and regulatory architecture of m 6 A methylation in cancer and establishes a foundation for leveraging m 6 A as a new class of biomarkers and therapeutic targets.
利益披露 Disclosure
Y. Zhao, None..
K. Chen, None..
H. Chen, None..
Y. Song, None..
K. Mojumdar, None..
W. Liu, None..
S. Ting, None..
A. Semwal, None.