Leveraging cross-source heterogeneity to improve the performance of bulk gene expression deconvolution.
Leveraging cross-source heterogeneity to improve the performance of bulk gene expression deconvolution.
复制标题
利用跨源异质性来提高批量基因表达反卷积的性能。
DOI：
10.1101/2024.04.07.588458
复制
发表时间：
2024
期刊：
影响因子：
--
通讯作者：
Zhou,XinMaizie
中科院分区：
文献类型：
--
作者：
Shen,Wenjun;Liu,Cheng;Hu,Yunfei;Lei,Yuanfang;Wong,Hau-San;Wu,Si;Zhou,XinMaizie
批量转录组学技术的主要限制是个体测量通常包含来自多个细胞群体的贡献，阻碍了对患病组织内细胞异质性的鉴定。为了从现有的大量转录组数据中提取细胞见解，我们提出了CSsingle，这是一种新的方法，旨在使用scRNA-seq参考将大量数据准确地解卷积为一组预定义的细胞类型。通过使用不同的真实的数据集进行全面的基准评估和分析，我们揭示了现有方法中固有的系统偏差，源于细胞大小或文库大小的差异。我们广泛的实验表明，与领先的方法相比，CSsingle具有上级的准确性和鲁棒性，特别是在处理来自细胞大小明显不同的细胞类型的散装混合物时，以及处理从不同来源获得的散装和单细胞参考数据时。我们的工作为批量和scRNA-seq数据的综合分析提供了一种有效和强大的方法，促进了各种生物学和临床研究。
A main limitation of bulk transcriptomic technologies is that individual measurements normally contain contributions from multiple cell populations, impeding the identification of cellular heterogeneity within diseased tissues. To extract cellular insights from existing large cohorts of bulk transcriptomic data, we present CSsingle, a novel method designed to accurately deconvolve bulk data into a predefined set of cell types using a scRNA-seq reference. Through comprehensive benchmark evaluations and analyses using diverse real data sets, we reveal the systematic bias inherent in existing methods, stemming from differences in cell size or library size. Our extensive experiments demonstrate that CSsingle exhibits superior accuracy and robustness compared to leading methods, particularly when dealing with bulk mixtures originating from cell types of markedly different cell sizes, as well as when handling bulk and single-cell reference data obtained from diverse sources. Our work provides an efficient and robust methodology for the integrated analysis of bulk and scRNA-seq data, facilitating various biological and clinical studies.