`cytoftree`: extension of `cytometree` to analyze mass cytometry data

knitr::opts_chunk$set(tidy = TRUE)
knitr::knit_hooks$set(small.mar = function(before, options, envir) {
    if (before) par(mar = c(0, 0, 0, 0))  # no margin
})

Introduction to cytoftree

cytoftree is an extension to cytometree function to analyze mass cytometry data. These data are specific due to a high number of zero and the high number of markers (up to 100 potentially). cytoftree is based on cytometree's algorithm which is the construction of binary tree, whose nodes represents cell sub-populations, and slighly modified to take into account the specification of mass cytometry data.

Data transformation

According to the literature, mass cytometry data must be transform to get best partitions. We propose different transformations: asinh (as default), biexp, log10 or none (without transformation).

Binary tree construction

  1. At each node, for each marker, the cells with zero values are temporarily set aside from the other cells.

  2. The remaining observed cells (or "events") and markers are modeled by both a normal distribution (so unimodal), and a mixture of 2 normal distributions (so bimodal).

  3. If the AIC normalized differences $D$ are significant, the cells are split into 2 groups according to the bimodal distribution. Cells with low values are annotated - (no marker) while cells with high values are annotated + (with marker). The cells with zero values are also annotated - (no marker).

  4. The binary tree is constructed until the cells can no longer be split into 2 groups.

Post-hoc annotation

Given the unsupervised nature of the binary tree, some of the available markers may not be used to find the different cell populations present in a given sample. To recover a complete annotation, we defined, as a post processing procedure, an annotation method which allows the user to distinguish two (or three) expression levels per marker.

Influenza vaccine response dataset analysis with cytoftree

In this example, we will use an influenza vaccine response dataset (from ImmuneSpace), with 39 markers. To speed-up the computation, we sampled 10 000 cells from this dataset.

Data preparation

First, we can look the structure and the markers of the data.

library(cytometree)
data(IMdata)
dim(IMdata)
colnames(IMdata)

Then, we also check the proportion of zero for each marker, particularity of mass cytometry data.

zero_proportion <- apply(IMdata[,-c(1,2)], 
                         MARGIN = 2, 
                         FUN = function(x){round(prop.table(table(x==0))["TRUE"]*100,2)})
zero_proportion

CytofTree function

According to the available markers, a gating strategy may be considered. In this example, we have a gating strategy to conserve only viable cells by splitting on the following markers : DNA1, DNA2, Cell_length, Bead and Dead. This way, we can be as close as possible to manual gating. To do this, we have to force the markers with the force_first_marker option (semi-supervised gating).

Then, to improve the performance of automating gating, we decided to transform data with asinh transformation (default transformation). Then, we have to choose which markers should be transformed using num_col argument. The columns Time et Cell_length are not mass cytometry measure and shouldn't be transformed.

num_col <- c(3:ncol(IMdata))

tree <- CytofTree(M = IMdata,
                  minleaf = 1,
                  t = 0.1,
                  verbose = FALSE,
                  force_first_markers = c("(Ir191)Dd_DNA1",
                                          "(Ir193)Dd_DNA2",
                                          "Cell_length",
                                          "(Ce140)Dd_Bead",
                                          "(In115)Dd_Dead"),
                  transformation = "asinh",
                  num_col = num_col)

max(tree$labels)

High dimensional issues

Due to the high number of markers, cytoftree provides high number of sub-populations. minleaf value for the minimum of cells by sub-population and t threshold for the depth of the binary tree can be modified to get more or less sub-populations. The plot_graph function provides a look on the binary tree, but should be unreadable due to the high number of sub-populations.

Annotation function

The annotation function allows to recover the incomplete annotation on sub-populations. combinations option provides the complete annotation on each sub-population.

annot <- Annotation(tree, plot = FALSE, K2markers = colnames(IMdata))
annot$combinations[1:5,]

Due to the high number of sub-populations, it's recommended to use RetrievePops function which provide informations for particular sub-populations.

RetrievePops : providing informations for particular sub-populations

RetrievePops provides several informations on specific sub-populations, in particular the proportions and the sub-populations merged.

phenotypes <- list()
phenotypes[["CD4+"]] <- rbind(c("(Ir191)Dd_DNA1", 1), c("(Ir193)Dd_DNA2", 1), 
                              c("Cell_length", 0), c("(Ce140)Dd_Bead", 0), 
                              c("(In115)Dd_Dead", 0), c("(Sm154)Dd_CD14", 0), 
                              c("(Er166)Dd_CD33", 0), c("(Nd150)Dd_CD3", 1), 
                              c("(Nd143)Dd_CD4", 1))

phenotypes[["CD8+"]] <- rbind(c("(Ir191)Dd_DNA1", 1), c("(Ir193)Dd_DNA2", 1), 
                              c("Cell_length", 0), c("(Ce140)Dd_Bead", 0), 
                              c("(In115)Dd_Dead", 0), c("(Sm154)Dd_CD14", 0), 
                              c("(Er166)Dd_CD33", 0), c("(Nd150)Dd_CD3", 1), 
                              c("(Nd144)Dd_CD8", 1))

pheno_result <- RetrievePops(annot, phenotypes = phenotypes)

# CD4+
pheno_result$phenotypesinfo[[1]]

# CD8+
pheno_result$phenotypesinfo[[2]]

Proportions comparison between manual and automatic gating

We can compare proportions providing by automatic gating (cytoftree) and manual gating for the selected sub-populations.

automating <- c(pheno_result$phenotypesinfo[[1]]$proportion,
                pheno_result$phenotypesinfo[[2]]$proportion)
manual <- c(0.1824389, 0.06523925)
resu <- rbind(manual, automating)
rownames(resu) <- c("Manual Gating", "Automating Gating")
colnames(resu) <- c("CD4+", "CD8+")
knitr::kable(resu, digits = 3)

cytoftree provides good results, close to the proportions getting by manual gating.



Try the cytometree package in your browser

Any scripts or data that you put into this service are public.

cytometree documentation built on Dec. 5, 2019, 1:06 a.m.