In jlaffy/scrabble: Perform Exploratory Computational Analyses on Pre-Processed Single-Cell RNA-Sequencing Data

m = rowcenter(logtpm(bt771)) clusters = hca(m, clusters = T, min.cluster.size = 5, max.cluster.size = 5)

How much does dea slow down with sapply iterations? ... it actually looks much better than I expected; the increase in number of dea runs is more or less linear with increase in time, suggesting little effect of sapply on slowing down computation.

one = clusters[[1]] mbm = microbenchmark::microbenchmark( zero = dea(x = one, y = m, fc = 2, p = 0.05), times = 30 )

two = clusters[rep(1, 10)] mbm2 = microbenchmark::microbenchmark( ten = sapply(two, dea, y = m, fc = 2, p = 0.05, simplify = F), times = 30 )

three = clusters[rep(1, 100)] mbm3 = microbenchmark::microbenchmark( hundred = sapply(three, dea, y = m, fc = 2, p = 0.05, simplify = F), times = 30 )

data = do.call(rbind.data.frame, list(mbm, mbm2, mbm3)) data = data %>% dplyr::mutate(expr = factor(as.character(expr), levels = c('zero', 'ten', 'hundred'))) class(data) = c('microbenchmark', 'data.frame') ggplot2::autoplot(data) ggplot2::autoplot(data, log = T)

Relative speed of foldchange, ttest, wilcoxtest

x = clusters[[1]] mbm_dea_parts = microbenchmark( foldchange = foldchange(x, m, cutoff = NULL, is.log = T), ttest = ttest(x, m, cutoff = NULL, adjust.method = 'BH'), wilcoxtest = wilcoxtest(x, m, cutoff = NULL, adjust.method = 'BH'), times = 100 )

ggplot2::autoplot(mbm_dea_parts)

jlaffy/scrabble documentation built on Nov. 16, 2019, 7:56 a.m.