Installation instructions

[lgeistlinger]

For the installation instructions under https://ccb-hms.github.io/osca-workbench/, consider the following:

• dependency structure: the list of packages in the BiocManager::install command can be condensed based on the dependency structure between packages. For example, the following packages are dependencies of scater and are thus installed at the time scater is installed and thus do not need to be listed here: SingleCellExperiment, ggplot2, scuttle, BiocNeighbors, BiocParallel, BiocSingular, pheatmap.
• remote packages: SeuratData is only available from github, change to lgeistlinger/SeuratData
• hardware requirements: having the Ncpus = 4 argument in the recommended/default installation command seems problematic as many users in less privileged parts of the world might stumble over this one. My suggestion would be to provide the installation command without the Ncpus argument and have a note below that points out that the installation process can be expedited on multi-core machines with the Ncpus argument.
• package updates: the recommended approach is to use the latest release version of all packages. I understand that, in practice, this note is often met with a somewhat lazy on-demand response, but as in driving school we have to be clear what the guidelines are. Based on previous workshops, the number one reason for issues in workshops are outdated versions of R, Bioconductor, and individual packages.

[lgeistlinger]

let's also follow the approach of the bioc-project and bioc-rnaseq modules, and refer for platform-specific R and RStudio installation instructions to the installation instructions provided on the landing page of the bioc-intro module as opposed to having a copy of them on here.

[andrewGhazi]

1. Okay, I'll look through the dependency trees to condense the install command.
2. Maybe I should just add devtools::install_github('satijalab/seurat-data') ?
3. I tested it and BiocManager::install() will proceed without error if Ncpus is greater than the number of available cores.
4. Okay, I'll remove the note about updates. I've run into the complementary problem in the past, where students get stuck at the outset because they are trying to update hundreds of packages. I guess there's no perfect solution...

[lgeistlinger]

Maybe I should just add devtools::install_github('satijalab/seurat-data') ?

BiocManager::install invokes remotes::install_github when provided a github account and repo string, that means here: BiocManager::install("satijalab/seurat-data"). That will help to keep the installation instructions simple by providing a single installation command.

For the end user using "satijalab/seurat-data" should be fine and likely better to use here, but note that there have been issues in the past with repo string ("seurat-data") not matching the package name ("SeuratData") in GHA installations as part of the github page build. "lgeistlinger/SeuratData" circumvents this problem.

Okay, I'll remove the note about updates. I've run into the complementary problem in the past, where students get stuck at the outset because they are trying to update hundreds of packages. I guess there's no perfect solution...

If we use personal setups for the course then it needs to be communicated that the latest release version of R, Bioconductor, and individual packages should be used. Otherwise instantiated cloud setups as in workshop.bioconductor.org serve such instances to the participants which should minimize such issues.

[andrewGhazi]

Okay, I addressed these items in #55