inst/scripts/make-data.R

In julia-wrobel/VectraPolarisData: Vectra Polaris and Vectra 3 multiplex single-cell imaging data

# function define below converts vectra .txt files to spatial experiment object

# Loads data from a Vectra3 or Vectra Polaris experiment in cell segmented tabular form
# the function right now assumes a file structure where folder contains multiple txt files, each is from one image
# can also handle "Consolidated_data" file of multiple images/subjects


## sample_path should point to a directory with one or multiple txt files
readVectraTable <- function(sample_path = "",
                            save = FALSE){

    # get file names
    # throw errors/warnings
    files <- list.files(sample_path, pattern = ".txt",
                        full.names = TRUE)


    # for each individual txt file, do the following
    spel <- lapply(seq_along(files), function(f){
        # load data
        if(require(phenoptr)){df <- read_cell_seg_data(files[f])
        }else{
            library(vroom)
            df <- vroom(files[f], col_names = TRUE)
            colnames(df) <- gsub(pattern = " \\(Normalized Counts, Total Weighting\\)", "",
                                 colnames(df))
        }

        # convert column names to snake case
        df = janitor::clean_names(df)

        # define matrices for assays. All assays have same rownames
        intensities_assay <- t(as.matrix(subset(df, select = names(df)[intersect(grep("entire_cell", names(df)), grep("_mean", names(df)))])))
        rownames(intensities_assay) <- substr(rownames(intensities_assay), start = 13, stop = nchar(rownames(intensities_assay))-5)
        rownames(intensities_assay)[grep("dapi", rownames(intensities_assay))] <- "dapi"

        nucleus_assay <- t(as.matrix(subset(df, select = names(df)[intersect(grep("nucleus", names(df)), grep("_mean", names(df)))])))
        rownames(nucleus_assay) <- substr(rownames(nucleus_assay), start = 9, stop = nchar(rownames(nucleus_assay))-5)
        rownames(nucleus_assay)[grep("dapi", rownames(nucleus_assay))] <- "dapi"

        membrane_assay <- t(as.matrix(subset(df, select = names(df)[intersect(grep("membrane", names(df)), grep("_mean", names(df)))])))
        rownames(membrane_assay) <- substr(rownames(membrane_assay), start = 10, stop = nchar(rownames(membrane_assay))-5)
        rownames(membrane_assay)[grep("dapi", rownames(membrane_assay))] <- "dapi"


        ###### add check that each assay has same number of variables
        ###### add check that variable names are in the same order

        ## define variables for different slots
        # stat summaries of marker intensities including standard deviation, min, max, and total go into colData
        colData_vars <- c("cell_id", "tissue_category", "slide_id",
                          "cell_x_position", "cell_y_position",
                          names(df)[grep("area|phenotype|axis|compactness|min|max|std_dev|total", names(df))] )

        # make into spe object
        SpatialExperiment(
            assays = list(intensities = intensities_assay,
                          nucleus_intensities = nucleus_assay,
                          membrane_intensities = membrane_assay),
            sample_id = df$sample_name,
            colData = DataFrame(subset(df, select = colData_vars)),
            spatialCoordsNames = c("cell_x_position", "cell_y_position")
        )
    }) # end lapply
    spe <- do.call(cbind, spel)

    # save or return data
    if (save) {
        save(spe, file = out_path)
    }

    return(spe)
}