Introduction

FANTOM Consortium 2014. A promoter-level mammalian expression atlas. Nature 507, 462-470, 2014. PMID: 24670764

Exercise

Figure S2d:

The autors classified promoters based on their initiation pattern into two classes: broad and sharp. They then studied the base composition around them. They found a characteristic signal in the region +50 to +200 bp from the TSS that show a strong periodic signal. This signal is similar to the one from MNase-seq experiments of region with high nucleosome affinity.

Hints and recipes

• First classify promoters based on their initiation patterns. To do so, use CAGE data from the ENCODE consortium (all samples cell longPolyA).
  • To do so you have first to extract CAGE data around EPDnew promoters using ChIP-Extract. Use EPDnew 003 for hg19 as reference feature (oriented) and ENCODE CAGE data (all samples cell longPolyA) as target feature (only the plus strand). Select a range from -100 to 100, Window width 1 and count cut-off 9999999.
  • Run the job and save the sga file as "promoters.sga" and the TEXT file as "promoters_expression.out".
  • Now evaluate promoter initiation pattern in R. Peak promoters have CAGE distribution very close to the TSS whereas broad promoters have CAGE tags spread around a larger area. You can use the Dispersion Index [Note: this is not the exact code used by the autors, is our own method but should provide similar results] in R using the following code:
    dispersion <- function(x){
  tss <- ceiling(length(x)/2)
  region <- length(x)
  if(sum(x) == 0) {
    m <- 0
    s <- -1
  } else {
    m <- sum(x*(1:region))/sum(x)
    s <- (sum((x*(1:region-m)**2))/sum(x))**0.5
  }
  return(s)
}

p.expression <- read.table("promoters_expression.out")
p.dispersion <- apply(p.expression, 1, dispersion)

# The following read the SGA file with the promoter collection
promoters <- read.table("promoters.sga", as.is=T)

# The following concatenate the SGA file with the Dispersion Index
promoters <- cbind(promoters,p.dispersion)

# The following write out a new SGA file with 7th column as Dispersion Index
write.table(promoters, file="promotersDispersionIndex.sga", quote=F, sep="\t", col.names=F, row.names=F)

    In R check the distribution of the Dispersion Index values ad try to set a cut-off between broad and peak promoters (peak promoters should be 10-15% of all active promoters). [Note: some promoters have no expression, the Dispersion Index is set to -1, exclude those from the analysis]. To do so, you can use the following plot:
    plot(density(p.dispersion), type="l")

  • Use a cut-off value of 12 to group promoters using the following code:
    peak <- which(p.dispersion > 0 & p.dispersion < 12)
broad <- which(p.dispersion >= 12)
write.table(promoters[peak,], file="peakPromoters.sga", quote=F, sep="\t", col.names=F, row.names=F)
write.table(promoters[broad,], file="broadPromoters.sga", quote=F, sep="\t", col.names=F, row.names=F)


  • In your working directory you have two SGA file: "peakPromoters.sga" and "broadPromoters.sga" that you will now analyse for the presence of nucleosome signal.
• Convert final SGA files in FPS using ChIP-Convert tool (the SSA server that we have to use next only accepts FPS file format).
• Study dinucleotide frequencies around promoters using OProf tool from SSA. The autors used WW (W = A or T), but these are not the only dinucleotides that are know to have high nucleosome affinity. Try also SS (S = G or T), YY (Y = C or T) and RR (R = A or G). To do so, do as follow:
  • Upload your files as imput FPS file
  • Extend the range of the 3' border to 300, use a window size of 3 and a shift of 1 base
  • Use WW as consensus sequence with no mismatches and reference position 1
  • Run the job and save the text file
• Use R to import the text files and reproduce the figure.