DE dashboards

This provides instructions on how to generate the DE (and stimulation) dashboards using 2 methods:

Shiny version (see here)
flexdashboard version (see here)

Shiny version

Run the following scripts in the project directory to save the dashboard files:

	biodomain_correlation.R	biodomain_enrichment.R	save_dashboard_files.R	save_stim_dashboard_files.R
`-n` `--nameset`	Name of project	Name of project	Name of project	Name of project
`-g` `--genotype`	Name of genotype	Name of genotype	Name of genotype	Name of genotype
`-d` `--drug`	Name of drug	Name of drug	Name of drug	Name of drug
`-r` `--round-number`	Round of source file default: `R1`	Round of source file default: `R1`	Round of source file default: `R1`	Round of source file default: `R1`
`-t` `--type`	Cell type if single-cell data, otherwise `bulk` for bulk data or `stim` for stimulation analysis default: `bulk`	Cell type if single-cell data, otherwise `bulk` for bulk data or `stim` for stimulation analysis default: `bulk`	Cell type if single-cell data, otherwise `bulk` for bulk data default: `bulk`	--
`-a` `--analyses`	Name of comparisons (ORDER BY geno, drug, geno + drug OR stim in wtv, tgv, tgd) If `-t` is NOT `stim`, default: `Tg-VvsWt-V,Tg-DvsTg-V,Tg-DvsWt-V` If `-t` is `stim`, default: `Wt-StvsWt-Un,Tg-StvsTg-Un,Tg-D-StvsTg-D-Un`	Name of comparisons (ORDER BY geno, drug, geno + drug OR stim in wtv, tgv, tgd) If `-t` is NOT `stim`, default: `Tg-VvsWt-V,Tg-DvsTg-V,Tg-DvsWt-V` If `-t` is `stim`, default: `Wt-StvsWt-Un,Tg-StvsTg-Un,Tg-D-StvsTg-D-Un`	Name of comparisons (ORDER BY geno, drug, geno + drug) default: `Tg-VvsWt-V,Tg-DvsTg-V,Tg-DvsWt-V`	Name of comparisons (ORDER BY stim in wtv, tgv, tgd) default: `Wt-StvsWt-Un,Tg-StvsTg-Un,Tg-D-StvsTg-D-Un`
`-s` `--shrunken`	Flag indicating shrunken LFC is being used default: `FALSE` Note: This is more for consistency w/ the other scripts and to use default `-c` value of `shrunkenL2FC`. If providing custom shrunken LFC column name, including `-s` or not technically does not make a difference.	Flag indicating shrunken LFC is being used default: `FALSE`	Flag indicating shrunken LFC is being used default: `FALSE`	Flag indicating shrunken LFC is being used default: `FALSE`
`-u` `--unadjusted-p`	--	Flag indicating unadjusted (nominal) p-value is being used for filtering instead of adjusted p-value (in the case of unshrunken LFC) default: `FALSE` _Note: This is more for consistency w/ save_dashboard_files.R and to use default `-c` value for p-value filter. If providing custom p-value column name, including `-u` or not technically does not make a difference._	Flag indicating unadjusted (nominal) p-value is being used for filtering instead of adjusted p-value (in the case of unshrunken LFC) default: `FALSE`	--
`-i` `--gene-col`	Column name for gene ID (bulk) or symbol (single-cell) default: `GENE`	Column name for gene ID (bulk) or symbol (single-cell) default: `GENE`	Column name for gene ID (bulk) or symbol (single-cell) default: `GENE`	Column name for gene ID default: `GENE`
`-c` `--de-columns`	Column name for log fold change (can provide shrunken or unshrunken) If `-s`, default: `shrunkenL2FC` If `-t` is `bulk`/`stim`, default: `log2FoldChange` If `-t` is single-cell, default: `avg_log2FC`	If shrunken (and not stim), column name for shrunken log fold change OR if unshrunken (or stim), column names for log fold change and p-value column for filtering If `-s` and `-t` is NOT stim, default: `shrunkenL2FC` If `-s` and `-t` is stim, default: `shrunkenL2FC,padj` If `-t` is `bulk`/`stim`, default: `log2FoldChange,padj` OR `log2FoldChange,pvalue` if `-u` If `-t` is single-cell, default: `avg_log2FC,p_val_adj` OR `avg_log2FC,p_val` if `-u` Note: Must use padj < 0.05 threshold w/ either shrunken or unshrunken LFC for stim dashboard.	If shrunken, column names in the order of `sLFC,sval,LFC,pval,padj` OR if unshrunken, column names in the order of `LFC,p1,p2,(pct1,pct2)`, where `p1` is the same p-value column used for filtering in biodomain_enrichment.R, `p2` is the other (out of p-value and p-adj columns), and `pct1`/`pct2` only applies for single-cell data If `-s`, default: `shrunkenL2FC,svalue,log2FoldChange,pvalue,padj` If `-t` is `bulk`/`stim`, default: `log2FoldChange,padj,pvalue` OR `log2FoldChange,pvalue,padj` if `-u` If `-t` is single-cell, default: `avg_log2FC,p_val_adj,p_val,pct.1,pct.2` OR `avg_log2FC,p_val,p_val_adj,pct.1,pct.2` if `-u`	If shrunken, column names in the order of `sLFC,padj,sval,LFC,pval` OR if unshrunken, column names in the order of `LFC,adj,pval` If `-s`, default: `shrunkenL2FC,padj,svalue,log2FoldChange,pvalue` If NOT `-s`, default: `log2FoldChange,padj,pvalue`
`-l` `--lfc-threshold`	--	LFC threshold used to filter genes default: `log2(1.1)`	LFC threshold used to filter genes default: `log2(1.1)`	LFC threshold used to filter genes default: `log2(1.1)`
`-p` `--p-threshold`	--	P-value threshold used to filter genes default: `0.05` Note: Do not change default for stim dashboard.	P-value threshold used to filter genes default: `0.05`	--

Note: ensembl_ver and gProfiler term_size_limit are defined at the top of the scripts. The Ensembl genes data is saved inside reference/biodomains/ by the genes_info.R script located there.

Deploying to shinyapps.io

In the project directory, add a copy of app.R or app_stim.R (must be renamed to app.R if deploying) and www/.

Inside app.R, define the following variables at the top of the file:

nameset (for single-cell data only)
round_num
is_sc (for DE dashboard only)

Also, modify the following sections accordingly:

Project dropdown selection choices defined in projs near the top of the file
Dashboard title and authors inside dashboardHeader()
Uncomment "Wt drug effect" sidebar text if applicable inside output$page_legend

In www/style.css, uncomment the appropriate sections as indicated within the file:

For bulk data, uncomment rule to hide sidebar tabs
For single-cell data, uncomment rules to format sidebar options
For DE dashboards, uncomment rule for formatting L2FC correlation tab
For stimulation dashboards, uncomment rules for formatting L2FC correlation and GO terms enrichment tabs

See deploy.R for template of deploying to shinyapps.io. Make sure to add this file to .gitignore if using as it contains credential information.

If running into error regarding package version while deploying, see here to downgrade package version.

If published dashboard is having trouble loading, you may need to increase the instance size from the shinyapps.io dashboard: Settings > General > Instance Size

`flexdashboard` version

Use rmarkdown::render() to knit the dashboard.Rmd template with the following options:

base_dir: Path to project directory
- Must contain subfolder(s) named by typeout for single-cell data or nameset for bulk data
- Each subfolder must contain the respective DE results files named as: R#.nameset.(typeout).RNA.analyses.csv.gz
templates_dir: Path to templates directory (default: templates)
ensembl_genes: Data table containing Ensembl genes + modules info
- Can be obtained by calling get_genes_info() from genes_info.R
nameset: Name of project (e.g., PS19_C31)
geno: Name of genotype (e.g., PS19)
drug: Name of drug (e.g., C31)
round_num: Round of source file (default: R1)
typeout: Cell type if single-cell data
- If unset, nameset is used
analyses: Name of comparisons (default: c('Tg-VvsWt-V', 'Tg-DvsWt-V', 'Tg-DvsTg-V'))
- Provide as a vector of comparisons in the following order: geno, geno + drug (stim), drug (stim)
- Optionally, include a 4th comparison for drug (stim) effect in wildtype
- Specifies how the DE results files are named (see above)
de_cols: Column names for gene ID, log fold change, p-value, followed by any additional columns of interest (default: c('GENE', 'avg_log2FC', 'p_val_adj', 'pct.1', 'pct.2'))
- Log fold change will be used for ranking genes in gProfiler input and determining gene category
- P-value will be used for defining significance when determining gene category
de_names: Column labels for de_cols not including gene ID (default: c('Log2FC', 'Pval (adj)', 'Pct 1', 'Pct 2'))
pval_col: Column name for nominal p-value (default: p_val)
- Used in the L2FC correlation tab
lfc_unshrunken_col: Column name for unshrunken log fold change (default: avg_log2FC)
- Used in the L2FC correlation tab
fdr_cutoff: Significance value threshold (default: 0.05)
- Used for p-value specified in de_cols to determine gene category
lfc_cutoff_geno: Absolute genotype log fold change cutoff (default: 0.55)
- Used for displaying gene label in the L2FC correlation tab
lfc_cutoff_drug: Absolute drug log fold change cutoff (default: 0.5)
- Used for displaying gene label in the L2FC correlation tab

Note: Required options are in bold.

Troubleshooting

Load R and run script to generate dashboards:

$ ml R/4.0
$ Rscript scripts/PS19_C31.R

Ensure tidyr is updated to at least v.1.2.0 to use the names_vary argument for pivot_wider():

$ R
> install.packages('tidyr')

If running into ReferenceError: FlexDashboard is not defined, make sure pandoc is up-to-date. This can be done by specifying the RSTUDIO_PANDOC environment variable in ~/.bash_profile:

export RSTUDIO_PANDOC=/usr/lib/rstudio-server/bin/quarto/bin

To encrypt the HTML files using staticrypt, install the CLI locally with npm:

$ mkdir -p .npm/node_modules
$ npm install --prefix .npm/ staticrypt

Make sure to append to the PATH environment variable in ~/.bash_profile:

PATH=$PATH:$HOME/.npm/node_modules/.bin
export PATH

Encrypt all HTML files in the output directory, making sure to purge modules if there are conflicts:

$ module purge
$ find outputs/PS19_C31/ -type f -name "*.html" -exec staticrypt {} PASSPHRASE -o {} -r 1 \;

Alternatively, run gen_dashboards.sh and provide the name (-n) and password (-p) arguments:

$ sbatch gen_dashboards.sh -n PS19_C31 -p PASSPHRASE

Use

This repository is for internal use of the Longo Lab only. All work herein is under the exclusive copyright of the Longo Lab. Open source versions of the code may be made available in dedicated repositories associated with individual publications and projects at the discretion of the Longo Lab.

Longo-Lab / de_dashboards

readme

DE dashboards

Shiny version

Deploying to shinyapps.io

`flexdashboard` version

Troubleshooting

Use

Longo-Lab / de_dashboards

readme

DE dashboards

Shiny version

Deploying to shinyapps.io

flexdashboard version

Troubleshooting

Use

`flexdashboard` version