How to ensure consistent clone sizes across different plots?

[eba28]

Hello,

I'm comparing control and disease in scRNA-Seq data with paired TCR-Seq (in a Seurat v4 object) and I would like to show how the disease (on the right) has more clones and they are more expanded than the control (on the left). The groups are colored by cell type. However, the scales for the clone sizes are so different that it becomes misleading. I read through all of your documentation and couldn't find a good way to standardize them. I've tried messing with clone_scale_factor, rad_scale_factor, etc. as well as getting into the guts of the plot object itself. Do you know if it might be possible to set the same scale so that the two subsets can be directly compared?

Thanks for your great package and your time!

[Qile0317]

Closing because we resolved this through email. Here is a summary of the solution I wrote:

The workaround would be to adjust the plot axes manually, while maintaining the exact same clone_scale_factor and rad_scale_factor. I suggest you use a ggplot theme to see ticks on the X and Y axes and then whichever plot has the smaller dimensions you should use ggplot::xlim() and ggplot::ylim() to adjust.

Ensure as well to save the plot to an exact size with ggplot2::ggsave() instead of just using a view to both ensure perfect circularity and consistent sizes. The underlying implementation uses ggforce::geom_circle() which will give the exact same circles everytime, however when ggplot builds the plot it essentially will stretch it to fit the screen, which you can see in the attachment.

After adjusting the x and y ranges of the plots, it would probably be a good idea to use AdjustAPOTC() to manually adjust the relative positions of some clusters so that the side by side makes more sense.

And lastly, you can manually adjust the legend sizes argument so its the same on both plots.

Sorry you have to go through this somewhat tedious process, but as APackOfTheClones is essentially a "visual Heuristic", so it inevitably requires quite a bit of fine tuning. I've actually thought of a few approaches to guide the relative positioning w/the reduction and other plots through optimizing the positioning function based on a cost function that rewards good relative outlines, but I don't quite have the time to do this.