In charles-plessy/CAGEr: Analysis of CAGE (Cap Analysis of Gene Expression) sequencing data for precise mapping of transcription start sites and promoterome mining

To accelerate plotting and reduce the file size of vector graphics created with the plotCorrelation2 method, points are deduplicated after reducing their precision to a smaller number of significant digits. After benchmarking the alternatives below, I chose the approach based on complex numbers, because their performance is similar to data.table or dplyr objects, and because these functions are available from standard R installations.

options(width=120)
knitr::opts_chunk$set(cache  = TRUE, cache.lazy = FALSE)
knitr::opts_knit$set(verbose = TRUE)

set.seed(1)
pois1 <- rpois(1e7, 0.3)
pois2 <- rpois(1e7, 0.3)

nbin1 <- rnbinom(1e7, mu = 2, size = .1)
nbin2 <- rnbinom(1e7, mu = 2, size = .1)

# Same with a bit of noise
x     <- nbin1 * runif(1e7)
y     <- nbin2 * runif(1e7)

sx    <- signif(x, 2)
sy    <- signif(y, 2)

DataFrame tables are coerced to data.frame objects by the duplicated.DataFrame function, therefore execution times are similar.

f.df       <- function(x,y) {
  df <- unique(data.frame(x = x, y = y))
  data.frame(df$x, df$y)}

f.tbl <- function(x,y) {
  tbl <- dplyr::distinct(dplyr::tibble(x = x, y = y))
  data.frame(tbl$x, tbl$y)}

f.dt       <- function(x,y) {
  dt <- unique(data.table::data.table(x = x, y = y))
  data.frame(dt$x, dt$y)}

f.cplx     <- function(x,y) {
  u <- unique(complex(real=x, im=y))
  data.frame(Re(u), Im(u))}

f.cplx.Rle <- function(x,y) {
  u <- unique(S4Vectors::Rle(complex(real=x, im=y)))
  data.frame(Re(u), Im(u))}

microbenchmark::microbenchmark(
  f.df       (pois1, pois2),
  f.dt       (pois1, pois2),
  f.tbl      (pois1, pois2),
  f.cplx     (pois1, pois2),
  f.cplx.Rle (pois1, pois2),  times = 10L)

microbenchmark::microbenchmark(
  f.df       (nbin1, nbin2),
  f.dt       (nbin1, nbin2),
  f.tbl      (nbin1, nbin2),
  f.cplx     (nbin1, nbin2),
  f.cplx.Rle (nbin1, nbin2),  times = 10L)

microbenchmark::microbenchmark(
  f.df       (x, y),
  f.dt       (x, y),
  f.tbl      (x, y),
  f.cplx     (x, y),
  f.cplx.Rle (x, y),  times = 10L)

microbenchmark::microbenchmark(
  f.df       (sx, sy),
  f.dt       (sx, sy),
  f.tbl      (sx, sy),
  f.cplx     (sx, sy),
  f.cplx.Rle (sx, sy),  times = 10L)

charles-plessy/CAGEr documentation built on Aug. 2, 2024, 4:35 p.m.