大量转录组学的计算机解卷积方法可以表征肿瘤微环境的细胞组成，量化与患者预后和治疗反应相关的细胞类型的丰度。虽然第一代反卷积方法依赖于预先计算的少数细胞类型的转录特征，但第二代方法可以使用单细胞数据进行训练，以解开更细粒度的细胞表型和状态。这些新方法还可以应用于空间转录组数据，以揭示肿瘤的空间组织。在这篇综述中，我们描述了最先进的反卷积方法（第一代、第二代和空间），这些方法可用于研究肿瘤微环境，讨论它们的优点和局限性。最后，我们对需要克服的挑战进行了展望，以释放下一代反卷积在肿瘤学和生命科学领域的全部潜力。版权所有 © 2024。由 Elsevier Inc. 出版。

Methods for in silico deconvolution of bulk transcriptomics can characterize the cellular composition of the tumor microenvironment, quantifying the abundance of cell types associated with patients' prognosis and response to therapy. While first-generation deconvolution methods rely on precomputed, transcriptional signatures of a handful of cell types, second-generation methods can be trained with single-cell data to disentangle more fine-grained cell phenotypes and states. These novel approaches can also be applied to spatial transcriptomic data to reveal the spatial organization of tumors. In this review, we describe state-of-the-art deconvolution methods (first-generation, second-generation, and spatial) which can be used to investigate the tumor microenvironment, discussing their strengths and limitations. We conclude with an outlook on the challenges that need to be overcome to unlock the full potential of next-generation deconvolution for oncology and the life sciences.Copyright © 2024. Published by Elsevier Inc.