ccicb / crux Goto Github PK

CRUX currently has numerous dependencies - many of which are entirely unnecessary:

• utilitybeltfmt - replace with selkamand/assertions
• utilitybeltgg - leave for now but consider building functionality into CRUX later
• fs: only used with shinyfiles and can be removed once we replace shinyfiles with shinyapps compatible solutions
  htmltools: used only in function with_red_star - which is never used. Can delete function and dependency
• readr: used only in a few locations, can replace with data.table::fread or utils::read.csv
• Pheatmap: moved to suggests
• tibble: can just call dplyr::tibble(). tibble doesn't need to be in description file. UPDATE: we use tibble::column_to_rownames and since we have a tibble dependency indirectly via dplyr even if we get rid of it - theres no point removing unless we drop dplyr dependency

Essentially all text in the app, but particularly within the plots - is far too small by default

outputOptions(..., suspendWhenHidden = FALSE) has been used in several places.

This causes code to run at times thats unnecessary. Remove all instances of it.

If wanting to change UI based on server conditions. Use hidden tabsetpanels and updateTabsetPanels in server functions instead

Has lots of properties we don't need.

We just need the cohort metadata, a load function, a status to tell us if the data is loaded, and a slot to store loaded data.

For public data: function should stream data into loaded_data slot
For user data: function should instantly load and have no function for reloading.

Then if we need to tackle memory management, we can make sure to only ever unload datasets where the 'load' function is not null.

All the other slots are unnecessary and just make use of the class more confusing.

Overall we're stripping back from something that was leading towards being multiomic - to truly just being genomics focused

We can already export to CVCDAP by exporting MAFs (currently added as 'CRUX' export), but lets make an entry specifically for it

Since many programs may take MAF + clinical annotations we should add a general 'write MAF + clinical' function that the MAF object and filenames for the MAF & clinical TSV as arguments

For genepattern - consider exporting to IGV.js or GISTIC2

Landing page should describe CRUX in a few words, then ask whether they want to start exploring

1. 'Inbuilt TCGA datasets' or
2. 'Their own data'

depending on which button they click, it should take them to single cohort summary or the import data module

Its just confusing to have a table of included datasets with a 'not loaded' column early on in the UI - end users just end up trying to load the datasets of interest

Cohort comparisons using subsets from one group vs another group is a painpoint.

Would be useful to be able to create a subset & make another cohort from original minus subset

That way we can do subset vs original without including the same samples in both groups

Should be able to configure additional datasets to import via a deployment parameter.

See golem_opts

TCGA and PCAWG data can still be preloaded

Windows smartscreen blocks CRUX

Can fix by right clicking --> properties --> selecting unblock option at bottom of page but should investigate how to become a 'recognised' app on windows

If nothing else, we at least need to add the instructions on how to run CRUX into the manual

Create another shiny server that takes gistic output zips and creates a CRUX gistic object (Rds of maftools::readGistic)

Then a single file import can be used. Will mean we lose the option of viewing 'deep/shallow/all' peaks but could build presets for all versions for TCGA/PCAWG and make automatically available

are they deleted immediately after last use or only once instance shuts down.

Consider adding support for instantaneous imported file deletion on shinyapps only. we do NOT want to do this if its running locally. One way of checking would be to see if fileInput$datapath is in temp folder. Then the second we've read the data we can delete it

Add note for highly sensitive data, to download the R package / binaries and use CRUX locally

Its confusing that you currently wait for elements like the pickerinput that lets you select genes to pop up and theres nothing to indicate any loading of data is happening

Add 'save dataset' functionality that lets you save a maf_dataset_wrapper in an RDS file and implement some read functionality to quickly add it into your data pool

ShinyFiles browses the local environment (server filepath) NOT the users environment

Test with edge and multiple users

Investigate where to pull list of genes its possible to annotate GISTIC chromplot. Pulling from MAFS is not ideal since its the GISTIC data that likely contains annotations

by default sorts are based on frequency but once you select a custom set of genes sort order is based on alphabetic/reverse alphabetic order. When selecting custom set of genes I’d personally prefer them to appear in

1. the order they were selected OR
2. sort based on number of mutations.

Need to figure out if this behaviour is from CRUX or maftools

get CCICB/interchangeGUI live before CRUX goes out to allow Gistic ZIP conversion to maftools GISTIC object. Basic functionality is already in CCICB/interchange

maftools::gisticChromPlot(gistic = gistic(), fdrCutOff = input$in_num_fdrcutoff, :
Could not find mutations

Happens for all cohorts i've tested

Plots should appear above tables.
Consider Oncoplots First

Several long operations stop the user doing anything else, with no indication of when itll be finished (e.g. export of large cohorts for mutalisk).

It would be best in thse cases to produce a pop up that tells the user what the program is doing, and doesn't let them back into the main interface until the program is complete.

I need to figure out if this behaviour is possible then implement if it is

Error messages shown to user via sweet alerts (e.g. in mutational signatures module when incorrect module is given) are often cluttered by cli formatting chars that are useful when using an ID but don't render in html. add a function to strip these out

cli::test_that_cli should provide inspiration of way to do this (see config="plain")

Once function is added, wrap any error strings in a trycatch sweetalert with this function.

single sample analysis > THCA > default first sample > Clonal Heterogeneity tab > 'Error in loadNamespace(x): there is no package called 'mclust'

Should be automatically downloaded by maftools dependency chaining.

1. update maftools

if that fails, adding it as a CRUX dependency might be worthwhile as a short-term fix

See TCGAbiolinksGUI for an example

If running locally - save datasets generated by the user

maftools::loadTCGA should handle this for TCGA datasets

will need to create a loadPCAWG function to do this for PCAWG data. Eventually would be nice to have this function built into maftools, might be worth a pull request.

Must do this so package is light enough to upload to shinyapps

Probably a CRUX 2.0 feature

Questions we'd need to answer:

1. How do you group into cohorts?
2. Are MAF format mutation files readily available
3. What 'clinical annotations' should be included?

Survival Analysis

On PC (Windows10) bug survival analysis crashes

Use wellPanels, not boxes

Compare survival based on clinical metadata

An error has occurred. Check your logs or contact the app author for clarification.

Also sometimes seems to run twice when progressing through import wizard as normal

Merge module multi-input has some good formatting - lets just steal that

home -> house
warning -> ?
file-upload -> ?
Lots more

Last module where this is required

Select subset by sample, choose a sample, then with that sample selected, change the dataset.

Warning: Error in stringr::str_detect: `string` must be a vector, not a <simpleError/error/condition> object.
  [No stack trace available]
Closing App

Save time when doing things like isolating triple negative breast cancer cohort

Change warning message to venn diagram of sample IDs in mutalisk dataset vs MAF dataset