reduceDimension: Compute a projection of a CellDataSet object into a lower...

Description Usage Arguments Details Value


Monocle aims to learn how cells transition through a biological program of gene expression changes in an experiment. Each cell can be viewed as a point in a high-dimensional space, where each dimension describes the expression of a different gene in the genome. Identifying the program of gene expression changes is equivalent to learning a trajectory that the cells follow through this space. However, the more dimensions there are in the analysis, the harder the trajectory is to learn. Fortunately, many genes typically co-vary with one another, and so the dimensionality of the data can be reduced with a wide variety of different algorithms. Monocle provides two different algorithms for dimensionality reduction via reduceDimension. Both take a CellDataSet object and a number of dimensions allowed for the reduced space. You can also provide a model formula indicating some variables (e.g. batch ID or other technical factors) to "subtract" from the data so it doesn't contribute to the trajectory.


reduceDimension(cds, max_components = 2, reduction_method = c("DDRTree",
  "ICA", "tSNE", "SimplePPT", "L1-graph", "SGL-tree"), norm_method = c("log",
  "vstExprs", "none"), residualModelFormulaStr = NULL, pseudo_expr = 1,
  relative_expr = TRUE, auto_param_selection = TRUE, verbose = FALSE,
  scaling = TRUE, ...)



the CellDataSet upon which to perform this operation


the dimensionality of the reduced space


A character string specifying the algorithm to use for dimensionality reduction.


Determines how to transform expression values prior to reducing dimensionality


A model formula specifying the effects to subtract from the data before clustering.


amount to increase expression values before dimensionality reduction


When this argument is set to TRUE (default), we intend to convert the expression into a relative expression.


when this argument is set to TRUE (default), it will automatically calculate the proper value for the ncenter (number of centroids) parameters which will be passed into DDRTree call.


Whether to emit verbose output during dimensionality reduction


When this argument is set to TRUE (default), it will scale each gene before running trajectory reconstruction.


additional arguments to pass to the dimensionality reduction function


You can choose two different reduction algorithms: Independent Component Analysis (ICA) and Discriminative Dimensionality Reduction with Trees (DDRTree). The choice impacts numerous downstream analysis steps, including orderCells. Choosing ICA will execute the ordering procedure described in Trapnell and Cacchiarelli et al., which was implemented in Monocle version 1. DDRTree is a more recent manifold learning algorithm developed by Qi Mao and colleages. It is substantially more powerful, accurate, and robust for single-cell trajectory analysis than ICA, and is now the default method.

Often, experiments include cells from different batches or treatments. You can reduce the effects of these treatments by transforming the data with a linear model prior to dimensionality reduction. To do so, provide a model formula through residualModelFormulaStr.

Prior to reducing the dimensionality of the data, it usually helps to normalize it so that highly expressed or highly variable genes don't dominate the computation. reduceDimension() automatically transforms the data in one of several ways depending on the expressionFamily of the CellDataSet object. If the expressionFamily is negbinomial or negbinomial.size, the data are variance-stabilized. If the expressionFamily is Tobit, the data are adjusted by adding a pseudocount (of 1 by default) and then log-transformed. If you don't want any transformation at all, set norm_method to "none" and pseudo_expr to 0. This maybe useful for single-cell qPCR data, or data you've already transformed yourself in some way.


an updated CellDataSet object

monocle documentation built on Nov. 8, 2020, 5:06 p.m.