In RajLabMSSM/echoplot: echoverse module: Locus plot creation for fine-mapping and colocalization studies

pkg <- read.dcf("../DESCRIPTION", fields = "Package")[1]
library(pkg, character.only = TRUE)

library(`r pkg`)

Plotting loci with echoplot

echoplot contains various functions that can be used separately
from the comprehensive echolocatoR::finemap_loci() pipeline.

Generate a multi-view plot of a given locus using echoplot::plot_locus().

You can mix and match different tracks and annotations using the different arguments (see ?plot_locus for details).

The plot is centered on the lead/index SNP. If a list is supplied to zoom * plot_locus() returns a series of ggplot objects bound together with patchwork. One can further modify this object using ggplot2 functions like + theme(). + The modifications will be applied to all tracks at once.

Save a high-resolution versions the plot by setting save_plot=T.
Further increase resolution by adjusting the dpi argument (default=300).
Files are saved in jpg format by default, but users can specify their preferred file format (e.g. file_format="png")
Adjust the height and width of the saved plot using these respective arguments.
The plot will be automatically saved in the locus-specific directory as:
multiview__.jpg.

Load example data

Load example dataset of the results from fine-mapping the BST1 locus with finemap_loci(). Original data comes from the recent Nalls et al. (2019) Parkinson's disease GWAS (see ?BST1 for details).

library(ggplot2)
library(patchwork)

root.dir <- tempdir()
locus_dir <- file.path(root.dir,echodata::locus_dir)
dat <- echodata::BST1  
LD_matrix <- echodata::BST1_LD_matrix
LD_reference <- "UKB" # Used for naming saved plots
zoom = "10x"
show_plot <- FALSE

Full window

trk_plot <- echoplot::plot_locus(dat=dat, 
                                 LD_matrix=LD_matrix, 
                                 LD_reference=LD_reference,
                                 locus_dir=locus_dir,  
                                 save_plot=FALSE,
                                 show_plot=show_plot,
                                 zoom=zoom)

methods::show(trk_plot)

At multiple zooms

You can easily generate the same locus plot at multiple zoomed in views by supplying a list to zoom.
This list can be composed of zoom multipliers (e.g. c("1x", "2x")), window widths in units of basepairs (e.g. c(5000, 1500)), or a mixture of both (e.g. c("1x","4x", 5000, 2000)).
Each zoom view will be saved individually with its respective scale as the suffix (e.g. multiview.BST1.UKB.4x.jpg).
Each zoom view is stored as a named item within the returned list.

trk_zooms <- echoplot::plot_locus(dat=dat, 
                                  LD_matrix=LD_matrix, 
                                  LD_reference=LD_reference,
                                  locus_dir=locus_dir,  
                                  save_plot=FALSE,
                                  show_plot=show_plot,
                                  zoom = c("1x","5x","10x")) 
names(trk_zooms) # Get zoom view names

methods::show(trk_zooms)

Return as list

For even further control over each track of the multi-view plot, specify plot_locus(..., return_list=TRUE) to instead return a named list (nested within each zoom view list item) of ggplot objects which can each be modified individually.
Once you've made your modifications, you can then bind this list of plots back together with patchwork::wrap_plots(tracks_list, ncol = 1).

trk_plot_list <- echoplot::plot_locus(dat=dat, 
                                       LD_matrix=LD_matrix, 
                                       LD_reference=LD_reference,
                                       locus_dir=locus_dir,  
                                       save_plot=FALSE,
                                       show_plot=show_plot,
                                       zoom=zoom,
                                       return_list=TRUE)

view1_list <- trk_plot_list[[zoom]]
names(view1_list) # Get track names from a particular zoom view

Modify a specific tracks within a view.

# Modify your selected track
modified_track <- view1_list$GWAS + 
                      ggplot2::labs(title = "Modified GWAS") + 
                      ggplot2::theme_dark() +
                      ggplot2::theme(title = ggplot2::element_text(hjust = .5))
# Put it back into your track list
view1_list[["GWAS"]] <- modified_track
# Remove a plot you don't want
view1_list[["Genes"]] <- NULL
# Specify the relative heights of each track (make sure it matches your new # of plots!)
track_heights <- c(.3,.1,.3,1)

# Bind them together and plot
fused_plot <- patchwork::wrap_plots(view1_list, 
                                    heights = track_heights,
                                    ncol = 1)
methods::show(fused_plot)

Using XGR annotations

Whenever you use annotation arguments (e.g. xgr_libnames,Roadmap,nott_epigenome) the annotations that overlap with your locus will automatically be saved as GRanges objects in a locus-specific subdirectory:
results////annotation
If a selected annotation has previously been downloaded and stored for that locus, plot_locus() will automatically detect and import it to save time.

trk_plot.xgr <- echoplot::plot_locus(dat=dat, 
                                     LD_matrix=LD_matrix, 
                                     LD_reference=LD_reference,
                                     locus_dir=locus_dir, 
                               xgr_libnames=c("ENCODE_TFBS_ClusteredV3_CellTypes"), 
                               save_plot=FALSE,
                               show_plot=show_plot,
                               zoom=zoom)

methods::show(trk_plot.xgr)

Using Roadmap annotations

Using the Roadmap=T and roadmap_query="<query>" arguments searches the Roadmap for chromatin mark data across various cell-types, cell-lines and tissues.
Note that Roadmap queries requires tabix to be installed on your machine, or within a conda environment (conda_env = "echoR").
Parallelizing these queries across multiple thredas speeds up this process (nThread=<n_cores_available>), as does reusing previously stored data which is automatically saved to the locus-specific subfolder (<dataset_type>/<dataset_name>/<locus>/annotations/Roadmap.ChromatinMarks_CellTypes.RDS).

NOTE: Querying remote genomic data sources (e.g. Roadmap bigwigs) is not currently supported on Windows due to known bugs in rtracklayer.

if(.Platform!="windows"){
    trk_plot.roadmap <- echoplot::plot_locus(dat=dat, 
                                         LD_matrix=LD_matrix, 
                                         LD_reference=LD_reference,
                                         locus_dir=locus_dir,  

                                         roadmap=TRUE, 
                                         roadmap_query="monocyte", 

                                         save_plot=FALSE, 
                                         show_plot=show_plot,
                                         zoom="5x")
}

methods::show(trk_plot.roadmap)

Using Nott_2019 annotations

Query and plot brain cell type-specific epigenomic assays from Nott et al. (Science, 2019)
(see ?NOTT_2019.bigwig_metadata for details).

trk_plot.nott_2019 <- echoplot::plot_locus(dat=dat,  
                                           LD_matrix=LD_matrix, 
                                           LD_reference=LD_reference,
                                           locus_dir=locus_dir, 

                                           nott_epigenome=TRUE,  
                                           nott_binwidth = 200,
                                           nott_regulatory_rects = TRUE, 
                                           nott_show_placseq = TRUE,

                                           save_plot=FALSE,
                                           show_plot=show_plot,
                                           zoom=zoom)

methods::show(trk_plot.nott_2019)

Using QTL datasets

Plot multiple QTL p-value columns (or really P-value columns from any kind of dataset).
Each QTL dataset will be plotted as a new track.

dat1 <- data.table::copy(dat)
dat2 <- data.table::copy(dat)
# Make fake QTL P-values for the sake a demonstration
dat1$P <- abs(jitter(dat1$P, amount = 1e-15))
dat2$P <- abs(jitter(dat2$P, amount = 1e-16))
dat_ls <- list("fake_eQTL"=dat1, 
               "fake_sQTL"=dat2)

trk_plot.qtl <- echoplot::plot_locus_multi(dat_ls = dat_ls, 
                                           LD_ls = list(LD_matrix,LD_matrix),  
                                           locus_dir = locus_dir,
                                           show_plot = show_plot,
                                           zoom = "10x")