Validation of merge algorithm

Initialization

library(ggplot2)
library(ggrastr)
library(dplyr)
library(parallel)
library(dropestr)
library(dropEstAnalysis)
library(Matrix)

theme_set(theme_base)

kPlotsFolder <- '../../output/figures/'
kDataPath <- '../../data/dropest/'

dataset_names <- c('precise', 'unmerged')
kDropSeqPath <- paste0('dropseq/thousand/', c('est_01_16_precise/', 'est_01_16_unmerged/'), 
                       'thousand.rds')
kInDropPath <- paste0('SCG71/', c('est_11_16_poisson_simple/', 'est_11_16_unmerged/'), 
                      'SCG71.rds')
k10xPath <- paste0('10x/pbmc33k/', c('est_11_17_poisson/', 'est_11_17_unmerged/'), 
                   'pbmc33k_no_umis.rds')

kDataPaths <- list(dropseq=kDropSeqPath, indrop=kInDropPath, `10x`=k10xPath) %>% 
  lapply(function(l) paste0(kDataPath, l) %>% setNames(dataset_names))

holders <- mclapply(kDataPaths, function(x) mclapply(x, readRDS, mc.cores=length(x)),
                    mc.cores=length(kDataPaths))

validation_data <- lapply(holders, function(hs) list(cms=lapply(hs, `[[`, 'cm')))

validation_data$dropseq$cell_number <- 1000
validation_data$indrop$cell_number <- 5200
validation_data$`10x`$cell_number <- 30000

for (n in names(validation_data)) {
  cn <- validation_data[[n]]$cell_number
  validation_data[[n]]$umi_per_cb <- lapply(validation_data[[n]]$cms, function(cm) 
    sort(Matrix::colSums(cm), decreasing=T))
  validation_data[[n]]$real_cbs <- names(validation_data[[n]]$umi_per_cb$precise)[1:cn]
}

size_increase <- lapply(validation_data, function(d) (
  d$umi_per_cb$precise[d$real_cbs] - d$umi_per_cb$unmerged[d$real_cbs]) / 
    d$umi_per_cb$unmerged[d$real_cbs])

Figure

names(size_increase) <- c('Drop-seq, mixture', 'inDrop, BMCs', '10x, 33k PBMCs')
plot_df <- mapply(function(v, n) tibble(Increase=v, Dataset=n), size_increase, 
                  names(size_increase), SIMPLIFY=F) %>% bind_rows()

gg_size <- ggplot(plot_df) + 
  geom_histogram(aes(x=100 * Increase, fill=Dataset, y = 100 * 0.75 * ..density..), 
                 binwidth=0.75, color=alpha('black', 0.05), position='identity', alpha=0.5) +
  scale_x_continuous(expand=c(0, 0), limits=c(0.0, 20)) +
  scale_y_continuous(expand=c(0, 0), limits=c(0, 35)) +
  labs(x='Increase in #molecules per CB, %', y='Number of CBs, %') +
  theme_pdf(legend.pos=c(1, 1)) + theme(panel.grid.minor=element_blank())

ggsave(paste0(kPlotsFolder, "merge_size_increase.pdf"), gg_size, width=3.5, height=2.3)
gg_size

Session information