Improvements to plotting functions

[pcarbo]

@william-denault I'd like to suggest a few changes to the plotting functions that I think would make the plotting interface more logical and easier to use. I believe that these changes should be straightforward to implement.

First, I suggest having three functions:

plot_susiF
plot_susiF_pips
plot_susiF_effects

Also, please add a "plot" method which simply calls plot_susiF. See for example plot.ebnm in the ebnm package for an example of how to do this.

Second, the functions should not actually plot anything to the screen, but only return ggplot objects:

• plot_susiF_pips and plot_susiF_effects should each return a single ggplot object.

• plot_susiF should return a ggplot object or a list containing two ggplot objects. The output depends on an argument "which.plots" which looks like this (and will replace "pip_only"):

plot_susiF  =  function (..., which.plots = c("both", "pips", "effects"), ...)

Let me know if you have any questions about these suggested changes.

[pcarbo]

Also, plot_susiF should call the two other functions — right now it looks like plot_effect_susiF and plot_susiF have a lot of very similar or redundant code.

[pcarbo]

For plot_susiF_effects, there should be an options to plot the effects for all CSs (which is the default, e.g., effect = "all") as well as the option to plot multiple selected CSs (e.g., effect = 1:2).

[pcarbo]

You are using facet_grid for plotting mutliple effects?

[william-denault]

@pcarbo Yes I do use facet_grid for plotting mutliple effects.

I have updated the repo have you suggested let me know if this looks good to you. Thank you very for editing the vignette.

[pcarbo]

@william-denault A made a few improvements to the effect plot, and now I call it "plot_susiF_effect" to make the function name consistent with the other plotting functions.

A few notes:

• [x] Sadly, "aes_string" is deprecated in ggplot2. (I found it useful, but sadly we cannot use it anymore.)

• [x] There seems to be quite a bit of redundant ggplot code in this function; I improved this a bit, but perhaps you could improve more.

• [x] It looks like you are trying to add horizontal lines, but it doesn't show up in my effect plots?

• [x] You allow for additional arguments ("...") but it doesn't seem to be used anywhere? If so, you should remove it.

• [x] What does lfsr.curve = TRUE do? It doesn't seem to change my plot in the vignette example.

• [x] When I tried this I got an error:

plot_susiF_effect(fit,cred.band = FALSE,effect = "all")
# Error in rep(obj$outing_grid, indx_effect + 1) : invalid 'times' argument

• [x] I would expect these two lines to give me the same plot, but they gave me different plots:

plot_susiF_effect(fit,effect = 1:2)
plot_susiF_effect(fit,effect = 2:1)

[pcarbo]

Also note that "color" is the correct aesthetic in ggplot2, not "col".

[pcarbo]

I also made a few improvements to "plot_susiF_pip". (It was not an exported function, so I fixed that.)

Can you please add roxygen2 docs for this function?

Also, as before the additional arguments "..." seems to be unused? If so, I would remove it.

[william-denault]

Hi @pcarbo ,

I have added these improvement and your other suggestions.

The argument lfsr.curve = TRUE is for fsusie object that use the post_processing argument = "HMM" that I use in some other work. Essentially the output is the same but we postprocess the effect using an "ash-HMM" which allow to compute lfsr for each effect at each base pair. It is just another way to deal with the "incertainty"

You have a look at the example below

`library(ashr) library(wavethresh) set.seed(1)

Example using curves simulated under the Mixture normal per scale prior

rsnr <- 1 #expected root signal noise ratio N <- 100 #Number of individuals P <- 100 #Number of covariates/SNP pos1 <- 1 #Position of the causal covariate for effect 1 pos2 <- 5 #Position of the causal covariate for effect 2 lev_res <- 7#length of the molecular phenotype (2^lev_res) f1 <- simu_IBSS_per_level(lev_res )$sim_func#first effect f2 <- simu_IBSS_per_level(lev_res )$sim_func #second effect

G = matrix(sample(c(0, 1,2), size=N*P, replace=TRUE), nrow=N, ncol=P) #Genotype beta0 <- 0 beta1 <- 1 beta2 <- 1 noisy.data <- list()

for ( i in 1:N) { f1_obs <- f1 f2_obs <- f2 noise <- rnorm(length(f1), sd= (1/ rsnr ) var(f1)) noisy.data [[i]] <- beta1G[i,pos1]f1_obs + beta2G[i,pos2]f2_obs+ beta1G[i,15]*(f1_obs-f2_obs)+ noise

} noisy.data <- do.call(rbind, noisy.data)

Y <- noisy.data X <- G

Running fSuSiE

out <- susiF(Y,X,L=3 , prior = 'mixture_normal_per_scale') out2 <- susiF(Y,X,L=3 , prior = 'mixture_normal_per_scale', post_processing = "HMM") plot_susiF_effect(out) plot_susiF_effect(out2) plot_susiF_effect(out2, lfsr = FALSE) `

[pcarbo]

@william-denault Thanks, the effect plot function is looking better now. I made a few additional improvements; see my comment above with the checkmarks. I also streamlined the roxygen2 docs by putting the documentation for all plotting functions into a single help page; this will simplify the documentation.

I will keep this Issue open as I review the other plotting functions and continue to make improvements.

[pcarbo]

I added an option to plot_susiF_effect() to show the affected region, which is TRUE by default.

[pcarbo]

@william-denault I updated the plotting functions a bit more; please take a look at the new interface, which is illustrated in the new vignette.

[pcarbo]

@william-denault Can you fix plot_susiF_effect() so that the colors used always start with blue, then green, etc, regardless of how the "effect" argument is set? e.g., if I set effect = 7, the color should be the same as effect = 1. Does that make sense?