edawson / tidysig Goto Github PK

The default palettes, especially for signature activities, are insufficient. Bump to something like pals::alphabet (https://stackoverflow.com/questions/9563711/r-color-palettes-for-many-data-classes)

Rows: samples
Columns: Signatures
Fill: activity (normalized, 0-1)

A common use case is to compare both activities and signatures by group (e.g., by exposed status). The interface to do this is a little convoluted right now.

This will probably involve:

• a facet_grid (options: by_row, by_column)
• Label axis.strip.text with the group name

While it's currently possible to plot mutational context proportions, the API for doing so is bad. There should just be a plot_context wrapper to plot_signature().

Currently, only plot_activities works for both indels and SBS signatures. However, all of these functions could be generalized (i.e., calling a single "plot_signatures" function would handle both indels and SBS), which would greatly simplify the API.

It would be useful to add the site classification functions from presig to tidysig. This way, users would not have to switch programming languages. The classification algorithm of presig is pretty simple, so there's not a lot of work to accomplish this assuming we can get an easy-to-use VCF reader.

As is typical for ggplot2, ylim() truncates plot data rather than just "zooming" the axes.

To fix this, I need to use coord_cartesian(expand=FALSE, ylim=ylimits) rather than just ylimits.

Spread has been retired in favor of pivot_wider. The per_sample_*_summary functions rely on spread; this needs to be bumped to the new API.

When calling transform_sigprofiler_df, some CSV files will trigger warnings like this:

Transforming SigProfiler SBS96 signature / counts file.
Warning messages:
1: Trying to compute distinct() for variables not found in the data:
- `MutationType`
This is an error, but only a warning is raised for compatibility reasons.
The operation will return the input unchanged. 
2: Trying to compute distinct() for variables not found in the data:
- `MutationType`
This is an error, but only a warning is raised for compatibility reasons.
The operation will return the input unchanged.

This is likely because the variable munging process is still not completely robust to every version of SigProfiler dataframe observed. Hopefully, further testing will catch more of these. The best fix is likely to use something like janitor to automatically relabel the variables.

Given a tidysig dataframe, return the maximum activity value (i.e., max(Amount)) to set the y-limits automatically.