setup

• workflow was implemented and last executed successfully with
  R v4.4.1 with Bioc 3.20, and Python v3.11.3 with Snakemake v7.26.0
• R version and library have to be specified in the config.yaml file
  (e.g., R: "R_LIBS_USER=/path/to/library /path/to/R/executable")
• .Rprofile is used for handling and printing command line arguments
• logs/ capture .Rout files from R CMD BATCH executions
• data/ contains any synthetic and real data
• intermediate results are generated in outs/
• visualizations are generated in plts/

workflow

• <x> denotes a wildcard, namely: type, state, batch,
  simulation, score, selection, statistic,
  das = differential state analysis method

• 00-get_sim/dat.R

  • out: for simulations, data/sim/00-raw/t<t>,s<s>,b<b>.rds,
    for real data, data/dat/00-raw/<did>.rds (<did> = dataset identifier)
  • synthetic data generation (splatter::splatPopSimulate())
  • hereafter, t<t>,s<s>,b<b> = <sim>

• 01-pro_sim/dat.R

  • in: data/sim|dat/00-raw/<sim|dat>.rds
    out: data/sim|dat/01-fil/<sim|dat>.rds
  • minimal filtering keeping genes with count > 1
    in ≥ 10 cells, and cells with ≥ 10 detected genes
  • log-library size normalization (scater::logNormCounts())
  • highly variable gene (HVG) selection (scran::modelGeneVar())
  • principal component analysis (PCA) using HVGs (scater::runPCA())

• 02-sco.R

  • in: data/sim|dat/01-fil/<sim|dat>.rds
    out: outs/sim|dat/sco-<sim|dat>,<sco>.rds
  • source method from one of 02-sco-<sco>.R
  • compute gene-level metrics to quantify type-/state-specificity

• 03-sel.R

  • in: outs/sim|dat/sco-<sim|dat>,<sco>.rds
    out: outs/sim|dat/sel-<sim|dat>,<sco>.rds
  • source method from one of 03-sel-<sel>.R
  • select genes for reprocessing

• 04-rep.R

  • in: outs/sim|dat/sco-<sim|dat>,<sco>.rds
    out: data/sim|dat/02-rep/<sim|dat>,<sel>.rds
  • data reprocessing (PCA, clustering, reduction)

• 05-sta.R

  • in: data/sim|dat/02-rep/<sim|dat>,<sel>.rds
    out: outs/sim|dat/sta-<sim|dat>,<sel>,<sta>.rds
  • source method from on of 05-sta-<sta>.R
  • compute evaluation statistics

• 06-das.R

  • in: data/sim|dat/02-rep/<sim|dat>,<sel>.rds
    out: outs/sim|dat/das-<sim|dat>,<sel>,<das>.rds
  • source method from one of 06-das-<das>.R
  • perform differential state analysis (DSA)

• 07-eva.R

  • standalone script applied to experimental data only
  • collects results across all feature selection strategies,
    selects [10, 20, ..., 90%] for top-rank features, and recomputes
    evaluation statistics for accordingly reprocessed data (PCA, clustering)

• 08-plt_<out>-<plt>.R

  • in: outs/sim/<out>.rds
    out: plt/sim/<out>-<plt>.pdf
  • e.g., 08-plt_das-F1.pdf collects all DSA results
    (outs/sim/das-<sim>,<sel>,<das>.rds) and plots F1 scores
  • visualization of synthetic data analysis results

• 08-qlt_<out>-<qlt>.R

  • in: outs/dat/<out>.rds
    out: plt/dat/<out>-<qlt>.pdf
  • visualization of experimental data analysis results

• 09-aes.R

  • sourced to fix the order of feature scores (SCO),
    ground truth-based (DES) and other selections (SEL),
    and differential state analysis methods (DAS) across plots

• 10-session_info.R

  • lists and may be used to install all R packages used
    (across CRAN, GitHub, and Bioconductor), and writes the
    corresponding sessionInfo() output to session_info.txt