Supervised dimensionality reduction for exploration of single-cell data by HSS-LDA.
Supervised dimensionality reduction for exploration of single-cell data by HSS-LDA.
复制标题
DOI：
10.1016/j.patter.2022.100536
复制
发表时间：
2022-08-12
期刊：
影响因子：
6.5
通讯作者：
Bendall, Sean C.
中科院分区：
文献类型：
--
作者：
Amouzgar, Meelad;Glass, David R.;Baskar, Reema;Averbukh, Inna;Kimmey, Samuel C.;Tsai, Albert G.;Hartmann, Felix J.;Bendall, Sean C.
单细胞技术生成包含多种组学的大型高维数据集。降维捕获原始数据集的结构和异质性，创建有助于人类理解数据的低维可视化。现有算法通常是无监督的，使用测量的特征来生成流形，而忽略已知的生物标签，例如细胞类型或实验时间点。我们重新利用分类算法，即线性判别分析（LDA），用于单细胞数据的监督降维。 LDA 识别了预测因子的线性组合，可以最佳地分离先验类别，从而能够研究细胞异质性的特定方面。我们通过混合子集选择 (HSS) 来实现特征选择，并证明这种计算效率高的方法会生成非随机的、可解释的轴，适合不同的生物过程，例如随时间和细胞周期的分化。我们将 HSS-LDA 与几种流行的降维算法进行基准测试，并说明其在探索单细胞质谱流式分析、转录组学和染色质可及性数据方面的实用性和多功能性。 LDA 被重新用于单细胞数据的监督降维 HSS 提供额外的特征选择功能 性能与其他降维方法进行基准测试 LDA 可应用于任何单细胞组学数据 降维可以实现高维单细胞数据集的可视化。通常，这些算法是无监督的，不考虑已知的生物标签，例如细胞类型或实验时间点。 LDA 识别预测变量的线性组合，以最佳方式分离这些先验标签。结合 HSS 进行特征选择，LDA 生成可解释的轴，用于单细胞数据的可视化和分析。 LDA 可扩展到看不见的数据，也可以用作其他方法的输入，例如，集成多标签数据以进行 UMAP 可视化。通过定制轴来分离特定的、用户定义的标签，LDA 能够探索细胞异质性的不同方面，甚至在同一数据集中也是如此。更一般地说，这项工作引入了一种可视化单细胞数据的替代方法，并强调了可用于监督分析的丰富的实验元数据和细胞特征。阿穆兹加尔等人。重新利用分类算法 LDA，用于单细胞数据的监督降维。与 HSS 配合进行特征选择，LDA 可创建直观的可视化效果，将细胞与用户定义的标签（例如实验时间点和细胞周期阶段）分开。 LD 轴可以扩展到看不见的数据，并用作其他算法和方法的输入。 HSS-LDA 以几种流行的降维方法为基准，应用于多种生物问题和单细胞组学。
Single-cell technologies generate large, high-dimensional datasets encompassing a diversity of omics. Dimensionality reduction captures the structure and heterogeneity of the original dataset, creating low-dimensional visualizations that contribute to the human understanding of data. Existing algorithms are typically unsupervised, using measured features to generate manifolds, disregarding known biological labels such as cell type or experimental time point. We repurpose the classification algorithm, linear discriminant analysis (LDA), for supervised dimensionality reduction of single-cell data. LDA identifies linear combinations of predictors that optimally separate a priori classes, enabling the study of specific aspects of cellular heterogeneity. We implement feature selection by hybrid subset selection (HSS) and demonstrate that this computationally efficient approach generates non-stochastic, interpretable axes amenable to diverse biological processes such as differentiation over time and cell cycle. We benchmark HSS-LDA against several popular dimensionality-reduction algorithms and illustrate its utility and versatility for the exploration of single-cell mass cytometry, transcriptomics, and chromatin accessibility data. LDA is repurposed for supervised dimensionality reduction of single-cell data HSS provides additional feature selection functionality Performance is benchmarked against other dimensionality-reduction methods LDA can be applied to any single-cell omics data Dimensionality reduction enables the visualization of high-dimensional single-cell datasets. Typically, these algorithms are unsupervised, disregarding known biological labels such as cell type or experimental time point. LDA identifies linear combinations of predictors that optimally separate these a priori labels. Combined with HSS for feature selection, LDA generates interpretable axes for visualization and analysis of single-cell data. LDA is extensible to unseen data and can also be used as an input to other methods, for example, to integrate multi-label data for UMAP visualization. LDA enables the exploration of different aspects of cellular heterogeneity, even within the same dataset, by tailoring axes to separate specific, user-defined labels. More generally, this work introduces an alternative approach to visualize single-cell data and highlights the abundance of experimental metadata and cellular features that can be exploited for supervised analysis. Amouzgar et al. repurpose the classification algorithm, LDA, for supervised dimensionality reduction of single-cell data. Paired with HSS for feature selection, LDA creates intuitive visualizations, separating cells from user-defined labels such as experimental time point and cell-cycle phase. LD axes can be extended to unseen data and used as inputs for other algorithms and methods. HSS-LDA is benchmarked against several popular dimensionality-reduction methods and is applied to a diversity of biological problems and single-cell omics.
登录
查看更多内容
影响因子：
46.9
作者：
Han, Arnold;Glanville, Jacob;Hansmann, Leo;Davis, Mark M.
通讯作者：
Davis, Mark M.
昼夜节律正调节因子 CLOCK 和 BMAL1 通过 Cyclin B1 控制 G2/M 细胞周期转换
影响因子：
4.3
作者：
Farshadi, Elham;Yan, Jie;van der Horst, Gijsbertus T. J.
通讯作者：
van der Horst, Gijsbertus T. J.
DOI：
10.1126/science.1198704
发表时间：
2011-05-06
期刊：
Science (New York, N.Y.)
影响因子：
--
作者：
Bendall SC;Simonds EF;Qiu P;Amir el-AD;Krutzik PO;Finck R;Bruggner RV;Melamed R;Trejo A;Ornatsky OI;Balderas RS;Plevritis SK;Sachs K;Pe'er D;Tanner SD;Nolan GP
通讯作者：
Nolan GP
影响因子：
46.9
作者：
Becht, Etienne;McInnes, Leland;Newell, Evan W.
通讯作者：
Newell, Evan W.
影响因子：
46.9
作者：
Moon, Kevin R.;van Dijk, David;Krishnaswamy, Smita
通讯作者：
Krishnaswamy, Smita