R/cross_species.R

In whtns/seuratTools: Tools for Managing Seurat Objects

# you must set the rownames of the raw.data, data, and scale.data slots directly,
# as well as the the rownames of any gene loadings for PCAs you may have calculated.

#' Convert Seurat Objects from Mouse to Human
#'
#' @param seu Mouse seurat object
#'
#' @return
#' @export
#'
#' @examples
#'
#' convert_mouse_seu_to_human(baron2016singlecell)
convert_mouse_seu_to_human <- function(seu) {
    # transfer default species expression data to a species-specific assay
    seu[["mouse"]] <- seu[["gene"]]

    new_rownames <- convert_symbols_by_species(src_genes = rownames(seu), src_species = "mouse")

    seu_slots <- c("counts", "data", "scale.data", "meta.features")

    for (i in seu_slots) {
        current_slot <- slot(seu@assays[["gene"]], i)
        if (!(dim(current_slot) == c(0, 0))) {
            rownames(slot(seu@assays[["gene"]], i)) <- new_rownames
        }
    }

    return(seu)
}

#' Convert Seurat Objects from Human to Mouse
#' @param seu Human Seurat object
#' @param ... to be passed to \code{convert_symbols_by_species}
#'
#' @return
#' @export
#'
#' @examples
convert_human_seu_to_mouse <- function(seu, ...) {
    new_rownames <- convert_symbols_by_species(src_genes = rownames(seu), src_species = "human")

    seu_slots <- c("counts", "data", "scale.data", "meta.features")



    for (i in seu_slots) {
        rownames(slot(seu@assays[["gene"]], i)) <- new_rownames
    }

    return(seu)
}

#' Convert gene symbols between mouse and human
#'
#' @param src_genes Source gene symbol to be converted
#' @param src_species Source species
#'
#' @return
#' @export
#'
#' @examples
#'
#' convert_symbols_by_species("RXRG", "human")
#'
#' convert_symbols_by_species("Rxrg", "mouse")
#'
convert_symbols_by_species <- function(src_genes, src_species) {
    if (src_species == "human") {
        dest_species <- "mouse"

        dest_symbols <- src_genes %>%
            tibble::enframe("gene_index", "HGNC.symbol") %>%
            dplyr::left_join(human_to_mouse_homologs, by = "HGNC.symbol") %>%
            dplyr::distinct(HGNC.symbol, .keep_all = TRUE) %>%
            dplyr::mutate(MGI.symbol = dplyr::case_when(
                is.na(MGI.symbol) ~ stringr::str_to_sentence(HGNC.symbol),
                TRUE ~ MGI.symbol
            )) %>%
            dplyr::select(-gene_index) %>%
            identity()
    } else if (src_species == "mouse") {
        dest_species <- "human"

        dest_symbols <- src_genes %>%
            tibble::enframe("gene_index", "MGI.symbol") %>%
            dplyr::left_join(human_to_mouse_homologs, by = "MGI.symbol") %>%
            dplyr::distinct(MGI.symbol, .keep_all = TRUE) %>%
            dplyr::mutate(HGNC.symbol = dplyr::case_when(
                is.na(HGNC.symbol) ~ stringr::str_to_upper(MGI.symbol),
                TRUE ~ HGNC.symbol
            )) %>%
            dplyr::select(-gene_index) %>%
            # dplyr::mutate(HGNC.symbol = make.unique(HGNC.symbol)) %>%
            identity()
    }

    return(make.unique(dest_symbols[[2]]))
}

#' Integrate Seurat Objects from Mouse to Human
#'
#' @param mouse_seu_list Mouse Seurat object
#' @param human_seu_list Human Seurat object
#'
#' @return
#' @export
#'
#' @examples
#'
#' cross_species_integrate(list(baron2016singlecell = baron2016singlecell), list(panc8 = panc8))
#'
cross_species_integrate <- function(mouse_seu_list, human_seu_list, excluded_cells = NULL, ...) {
    mouse_seu_list <- purrr::map(mouse_seu_list, convert_mouse_seu_to_human)

    seu_list <- c(mouse_seu_list, human_seu_list)


    integrated_seu <- seurat_integrate(seu_list)

    # cluster merged seurat objects
    integrated_seu <- seurat_cluster(integrated_seu, resolution = seq(0.2, 2.0, by = 0.2))

    # add read count column
    integrated_seu <- add_read_count_col(integrated_seu)

    # annotate cell cycle scoring to seurat objects

    integrated_seu <- annotate_cell_cycle(integrated_seu, feature = "gene")

    # annotate excluded cells

    if (!is.null(excluded_cells)) {
        integrated_seu <- annotate_excluded(integrated_seu, excluded_cells)
    }

    # add marker genes to seurat objects

    integrated_seu <- find_all_markers(integrated_seu)

    return(integrated_seu)
}

#' Update human gene symbols in seurat object
#'
#' @param seu A Seurat object
#' @param assay Assay to use, Default = "gene"
#'
#' @return
#' @export
#'
#' @examples
update_human_gene_symbols <- function(seu, assay = "gene") {
    # browser()

    ensdb <- EnsDb.Hsapiens.v86::EnsDb.Hsapiens.v86

    symbols <- rownames(seu[[assay]])

    new_rownames <-
        AnnotationDbi::mapIds(ensdb, symbols, keytype = "SYMBOL", columns = c("SYMBOL", "GENEID")) %>%
        tibble::enframe("old_symbol", "ensgene")

    rownames(new_rownames) <- new_rownames$old_symbol

    seu[[assay]] <- Seurat::AddMetaData(seu[[assay]], new_rownames)

    new_rownames <-
        new_rownames %>%
        dplyr::left_join(annotables::grch38, by = "ensgene") %>%
        dplyr::distinct(old_symbol, .keep_all = TRUE) %>%
        dplyr::mutate(new_symbol = symbol) %>%
        dplyr::mutate(symbol = dplyr::coalesce(new_symbol, old_symbol)) %>%
        # tidyr::drop_na(symbol) %>%
        # dplyr::pull(symbol) %>%
        identity()

    seu_slots <- c("counts", "data", "scale.data", "meta.features")

    for (i in seu_slots) {
        if (length(slot(seu@assays[[assay]], i)) > 0) {
            rownames(slot(seu@assays[[assay]], i)) <- make.unique(new_rownames$symbol)
        }
    }

    variable_features <- VariableFeatures(seu[[assay]])
    if (length(variable_features) > 1) {
        new_variable_features <-
            dplyr::filter(new_rownames, old_symbol %in% variable_features) %>%
            dplyr::pull(symbol)

        VariableFeatures(seu[[assay]]) <- new_variable_features
    }

    return(seu)
}