Parse Kuono et al. data

In DeLorean: Estimates Pseudotimes for Single Cell Expression Data

---

title: Parse Kuono et al. data author: John Reid output: html_document

---

library(knitr)
library(rmarkdown)
#
# knitr options
#
opts_chunk$set(
    fig.path = 'figures/Kuono-',
    stop_on_error = TRUE,
    fig.width = 12.5,
    fig.height = 8)

library(rmarkdown)
library(devtools)
load_all('../..')
render('parse-Kouno.Rmd')

# suppressMessages(loadfonts())
#
# Stylesheet
#
options(markdown.HTML.stylesheet = system.file("inst/Rmd/foghorn.css",
                                               package="DeLorean"))

Kouno et al. investigated the transcriptional network controlling how THP-1 human myeloid monocytic leukemia cells differentiate into macrophages. Load their data from the CSV provided as supplementary information in their paper.

library(dplyr)
expr <- read.csv('../../data/gb-2013-14-10-r118-s3.csv')

Their cell labels have been messed up but Excel, so regenerate these.

time.points <- c("0h", "1h", "6h", "12h", "24h", "48h", "72h", "96h")
obs.times <- c(0, 1, 6, 12, 24, 48, 72, 96)
cell.ids <- (
    as.vector(
        sapply(1:length(time.points),
               function(t) sapply(1:120,
                                  function(c) sprintf("%s-%03d",
                                                      time.points[t],
                                                      c)))))
expr$cell <- factor(cell.ids, levels=cell.ids)
kouno.cell.meta <- data.frame(
    cell=expr$cell,
    capture=factor(
        as.vector(
            sapply(1:length(time.points),
                   function(t) rep(time.points[t], 120))),
        ordered=TRUE,
        levels=time.points)) %>% mutate(obstime=obs.times[capture])
kouno.gene.meta <- data.frame(gene=names(expr %>% select(-cell, -X)))

Kouno et al.'s data includes single cell expression values for r nrow(kouno.gene.meta) genes in r nrow(kouno.cell.meta) cells spread over r length(unique(kouno.cell.meta$capture)) time points.

Melt the data into a long format and log2 transform.

kouno.l <- melt(expr, variable.name="gene", value.name="expr")
kouno.l <- kouno.l %>% mutate(log.expr=log2(expr))

Plot the raw expression and the log transformed expression.

ggplot(kouno.l, aes(x=expr)) + geom_histogram()
ggplot(kouno.l, aes(x=log.expr)) + geom_histogram()

The log2 expression looks normally distributed. We will use this for as input to the DeLorean package. Cast the long format back into a matrix for use in the DeLorean package.

kouno.expr <- kouno.l %>% acast(gene ~ cell, value.var="log.expr")

Save the data in a format suitable for DeLorean package.

save(kouno.expr,
     kouno.gene.meta,
     kouno.cell.meta,
     file='../../data/KounoDeLorean.rda')

R version and packages used:

sessionInfo()

DeLorean documentation built on May 2, 2019, 9:24 a.m.