test_HPCell.R

# PREMISE
# these implementation of HPCL is highly experimental. 
# Therefore the interface, function names, arguments names, types of input 
# are subject to change and to be extended.
# At this stage, the input is a named list of Seurat objects. 
# In another branch (soon to be integrated), we accept SCE, H5 SCE, and anndata

BiocManager::install("MangiolaLaboratory/HPCell@grammar-hpc", upgrade_dependencies=FALSE)
BiocManager::install("crew", upgrade_dependencies=FALSE)
BiocManager::install("crew.cluster", upgrade_dependencies=FALSE)
BiocManager::install('satijalab/seurat-data', upgrade_dependencies=FALSE)


library(HPCell)
library(crew)
library(crew.cluster)
library(magrittr)
library(Seurat)
library(SeuratData)

options(Seurat.object.assay.version = "v5")
input_seurat <- 
    LoadData("pbmc3k") |>
    _[,1:500] |> 
    list() |>
    magrittr::set_names("pbmc3k")

# Define and execute the pipeline
input_seurat |> 
    
    # Initialise pipeline characteristics
    initialise_hpc(
        #debug_step = "empty_droplets_tbl_072a0a001eedb254",
        
        # Default resourced 
        computing_resources = crew_controller_local(workers = 10)
        
        # Slurm resources
        # computing_resources = 
        #   crew.cluster::crew_controller_slurm(
        #     slurm_memory_gigabytes_per_cpu = 5,
        #     workers = 50,
        #     tasks_max = 5,
        #     verbose = T
        #   )
    ) |> 
    
    # Remove empty outliers
    remove_empty_DropletUtils() |> 
    
    # Remove dead cells
    remove_dead_scuttle() |> 
    
    # Score cell cycle
    score_cell_cycle_seurat() |> 
    
    # Remove doublets
    remove_doublets_scDblFinder() |> 
    
    # Annotation
    annotate_cell_type() |> 
    
    normalise_abundance_seurat_SCT(factors_to_regress = c(
        "subsets_Mito_percent", 
        "subsets_Ribo_percent", 
        "G2M.Score"
    ))