Storing and Fine-Tuning APackOfTheClones Runs

The hardware and bandwidth for this mirror is donated by METANET, the Webhosting and Full Service-Cloud Provider.
If you wish to report a bug, or if you are interested in having us mirror your free-software or open-source project, please feel free to contact us at mirror[@]metanet.ch.

Introduction

As demonstrated in vignette("APackOfTheClones"), after processing the seurat & clonotype data properly with scRepertoire, vizAPOTC() provides a direct way to produce the ball-packing clonal expansion visualization, though for select users it may be somewhat clunky, if certain parameters need to be readjusted constantly. In this vignette, more details about how APackOfTheClones runs can be stored and re-adjusted will be covered - mainly through RunAPOTC(), APOTCPlot(), and AdjustAPOTC. Ensure to read the aforementioned vignette before this one.

As a reminder, here’s how to set up the seurat object and clonotype data

library(scRepertoire)

# A seurat object named `pbmc` is loaded with a corresponding `contig_list`
pbmc <- scRepertoire::combineExpression(
  scRepertoire::combineTCR(
    contig_list,
    samples = c("P17B", "P17L", "P18B", "P18L", "P19B", "P19L", "P20B", "P20L"),
    removeNA = FALSE,
    removeMulti = FALSE,
    filterMulti = FALSE
  ),
  pbmc,
  cloneCall = "gene",
  proportion = TRUE
)

print(pbmc)
#> An object of class Seurat 
#> 2000 features across 500 samples within 1 assay 
#> Active assay: RNA (2000 features, 2000 variable features)
#>  2 layers present: counts, data
#>  1 dimensional reduction calculated: umap

Overview

All of vizAPOTC()’s arguments are actually derived from RunAPOTC() and APOTCPlot(). The former is responsible for storing data of the S4 class ApotcData in the seurat object under a named list in @misc$APackOfTheClones under some character run ID, and the latter allows the visualization of these data objects with some customization. AdjustAPOTC has many arguments for adjusting the data associated with some APackOfTheClones run stored by RunAPOTC(), including adjusting cluster positions, colors, repulsion, etc. which can be visualized again with APOTCPlot().

Managing APackOfTheClones run data

RunAPOTC() has approximate the first half of vizAPOTC()’s arguments until max_repulsion_iter, meaning it has all the data subsetting, circle size scaling, and cluster repulsion capabilities covered in the previous vignette. The most essential difference is the presence of the argument run_id, which corresponds to an id for the ApotcData object associated with some run. If left blank, one will be automatically generated in the following format:

reduction_base;clonecall;keyword_arguments;extra_filter

where if keyword arguments and extra_filter are underscore (_) characters if there was no input for the ... and extra_filter parameters.

# Here is the function ran with its default parameters
pbmc <- RunAPOTC(pbmc)

#> Initializing APOTC run...
#> * Setting `clone_scale_factor` to 0.3
#> * id for this run: umap;CTstrict;_;_
#>
#> Packing clones into clusters
#> [==================================================] 100%
#>
#> repulsing all clusters | max iterations = 20
#> [==================================================] 100%
#>
#> Completed successfully, time elapsed: 0.155 seconds
#>

From the verbal queues, one can see how the run_id was set. Here’s it ran again but with more optional arguments and a custom run_id:

pbmc <- RunAPOTC(
    pbmc, run_id = "sample17", orig.ident = c("P17B", "P17L"), verbose = FALSE
)

Important Note

Practically speaking, most users probably only really need to ever work with visualizations from one run and mainly want to just adjust its parameters or the aesthetics of the plot - or they already will have several subsetted seurat objects. In these cases, feel free to completely ignore any of the parameters related to run_id or any other filtering parameter, as all functions default to using the latest run if no identifying parameters are passed in.

Utilities for Managing APackOfTheClones Runs

It is to note that the data abstraction here with a run_id is intentional, and users should not manually touch any of the ApotcData objects with the seurat object unless they are extremely familiar with the latest internal implementation. Instead, here is a collection of functions that may be useful:

getApotcDataIds(pbmc) gets all current run_id’s, if any.
getLastApotcDataId(pbmc) gets the latest run_id, if any.
containsApotcRun(pbmc, run_id = "foo") returns whether a run_id exists in the seurat object.
renameApotcRun(pbmc, old_run_id = "foo", new_run_id = "bar") renames runs.
deleteApotcData(pbmc, run_id = "foo") deletes all data associated with a certain run_id.

APOTCPlot

To visualize stored APackOfTheClones runs, APOTCPlot() takes in a seurat object and the run_id. If no run_id is provided, it defaults to using the latest run. All other parameters are same as in the second half of vizAPOTC(). Although it is noteworthy that if the user had always relied on auto-generated run_id’s then APOTCPlot() also has these subsetting arguments:

And putting in identical arguments to generate the original ApotcData would work too, but this approach is less recommended as its a lot more (unnecessarily) verbose. Here is APOTCPlot() in action:

# Here, plots for samples 17 - 20 as seen in the previous vignette are made, where
# `orig.ident` is a custom column in the example data with levels corresponding to sample ids:
# ("P17B" "P17L" "P18B" "P18L" "P19B" "P19L" "P20B" "P20L").

pbmc <- RunAPOTC(
  pbmc, run_id = "P17", orig.ident = c("P17B", "P17L"), verbose = FALSE
)

pbmc <- RunAPOTC(
  pbmc, run_id = "P18", orig.ident = c("P18B", "P18L"), verbose = FALSE
)

pbmc <- RunAPOTC(
  pbmc, run_id = "P19", orig.ident = c("P19B", "P19L"), verbose = FALSE
)

pbmc <- RunAPOTC(
  pbmc, run_id = "P20", orig.ident = c("P20B", "P20L"), verbose = FALSE
)

cowplot::plot_grid(
  APOTCPlot(pbmc, run_id = "P17", retain_axis_scales = TRUE, add_size_legend = FALSE),
  APOTCPlot(pbmc, run_id = "P18", retain_axis_scales = TRUE, add_size_legend = FALSE),
  APOTCPlot(pbmc, run_id = "P19", retain_axis_scales = TRUE, add_size_legend = FALSE),
  APOTCPlot(pbmc, retain_axis_scales = TRUE, add_size_legend = FALSE), # defaults to latest
  labels = c("17", "18", "19", "20")
)
#> * generated ggplot object
#> * generated ggplot object
#> * generated ggplot object
#> * using the latest APackOfTheClones Run Data with object id: P20
#> * generated ggplot object

AdjustAPOTC

This function’s parameters help modify certain attributes about APackOfTheClones runs, and has the exact same first six parameters as APOTCPlot() for managing which run to modify. It also possesses the same four repulsion arguments in vizAPOTC() and RunAPOTC() if a run is to be repulsed again. See the function level documentation for the following parameters that can modify cluster locations, colors, and the adjustment of the clone_scale_factor and rad_scale_factor:

new_rad_scale_factor = NULL,
new_clone_scale_factor = NULL,
relocate_cluster = NULL,
relocation_coord = NULL,
nudge_cluster = NULL,
nudge_vector = NULL,
recolor_cluster = NULL,
new_color = NULL,
rename_label = NULL,
new_label = NULL,
relocate_label = NULL,
label_relocation_coord = NULL,
nudge_label = NULL,
label_nudge_vector = NULL,
verbose = TRUE

Tip to Speeding up Adjustments

While the vizAPOTC() and RunAPOTC() / APOTCPlot() functions to do APackOfTheClones runs and generate the ggplot object are relatively fast even for very large seurat objects, plot display times can get quite long the more circles there are. AdjustAPOTC is meant to be used in an incremental/iterative manner, where the user can adjust one aspect, inspect the plot, and adjust again if necessary, and repeat - the long plot display times may pose as a large inconvenience.

Setting the argument detail to FALSE when using APOTCPlot() or vizAPOTC() can help - this will plot entire clusters as one large circle - speeding up plotting times significantly. The details of the individual clonotype circles will be lost but the inspection of other aspects like label locations will be a much more pleasant process.

Usage Examples

Renaming Cluster Labels

After identification of the cell identity of each seurat cluster, it may be useful to rename each cluster label to the actual identity of the cells of each cluster. here is a minimal example of how this process is intended to be done:

# Do a run with just the first 4 seurat clusters, and rename labels
pbmc <- RunAPOTC(
    pbmc,
    run_id = "first_four",
    seurat_clusters = 1:4,
    verbose = FALSE
)

pbmc <- AdjustAPOTC(
    pbmc,
    run_id = "first_four",
    rename_label = 1:4,
    new_label = letters[1:4],
    verbose = FALSE
)

APOTCPlot(
    pbmc,
    run_id = "first_four",
    show_labels = TRUE,
    retain_axis_scales = TRUE
)
#> * generated ggplot object

Extra Cluster Repulsion

Another possibility is that the default repulsion may have left the clusters too close - here this is corrected - and the labels are moved as well. Note the use of the pipe operator may also be useful.

pbmc <- pbmc %>%
    RunAPOTC(run_id = "foo", verbose = FALSE) %>%
    AdjustAPOTC(
        run_id = "foo",
        repulse = TRUE,
        repulsion_threshold = 0.5,
        verbose = FALSE
    )

APOTCPlot(
    pbmc,
    show_labels = TRUE,
    retain_axis_scales = TRUE,
    add_size_legend = FALSE
)
#> * using the latest APackOfTheClones Run Data with object id: foo
#> * generated ggplot object

These binaries (installable software) and packages are in development.
They may not be fully stable and should be used with caution. We make no claims about them.