How to generate a figure like the Supplementary figure 4C in your paper?

[YiweiNiu]

Hi,

I would like to plot a figure like the Supplementary figure 4C in your bioRxiv paper? Could you please give me some guides to do that?

Thanks,
Yiwei

[teunbrand]

Hello,

That figure is the result of an ATA from a series of KO vs WT comparisons. Here is an example for the test data. We make a copy of the test data to show how to do it for multiple samples, and because it is a copy, the log2 fold change will show as 0.

library(GENOVA)

exp <- get_test_data("40k")
explist <- list(dat1 = exp, dat2 = exp)

# Call TADs with insulation score
ins <- insulation_score(explist)
tads <- call_TAD_insulation(ins, min_strength = 0)[[1]]

# Feed TADs to ATA
ata <- ATA(explist, tads)
visualise(ata, colour_lim = c(0, 600))

^{Created on 2021-07-08 by the reprex package (v1.0.0)}

[YiweiNiu]

Thank you for your quick reply!

I could run the ATA analysis successfully on my data. Thanks to the clear documentation of GENOVA!

But I do not know how to plot a figure like this, which is from the Supplementary figure 4C of your paper.

[teunbrand]

I think @robinweide might now the exact details, but I presume it would have been something like the following. Example assumes you've run the previous chunk posted above.

library(data.table)
library(ggplot2)

# Melt array
df <- data.table(
  x = as.vector(slice.index(ata$signal, 1)),
  y = as.vector(slice.index(ata$signal, 2)),
  sample = as.vector(slice.index(ata$signal, 3)),
  value = as.vector(ata$signal)
)
# Calculate distance (unit is arbitrary due to ATA)
df[, dist := x - y]
# Take averages per off-diagonal band
df <- df[, list(value = mean(value)), by = c("dist", "sample")]
# Split by sample
df <- split(df, df$sample)

# Adjust sample 2 a bit so that a difference will show
# Don't do this with real data!
df[[2]][, value := sqrt(value)]
# Recombine the two samples, if you have more, this needs to be repeated for all samples
df <- df[[1]][df[[2]], on = c("dist")]
# Convert distance to TAD units
df[, dist := scales::rescale(dist, to = c(-2, 2))]

# Plot stuff
ggplot(df, aes(dist, value / i.value)) +
  geom_line(aes(colour = sample))

[YiweiNiu]

Bravo! Thank you very much. Now I understand the structure of ATA$signal.