# Load necessary libraries
pacman::p_load(Seurat, dplyr, ggplot2, ComplexHeatmap)

# Extract Marker Genes ----------------------------------------------------
{
    # Load the Seurat object
    seuObj <- readRDS("your_seurat_object.Rds")
    # Preprocess the data
    seuObj <- seuObj %>% NormalizeData() %>% FindVariableFeatures() %>% ScaleData()
    # Load marker genes and filter based on criteria
    marker_genes <- readr::read_csv("marker_genes.csv") %>% filter(avg_log2FC > 0.2, p_val_adj <= 0.01)
    # Extract markers
    topGenes <- marker_genes %>% filter(cluster == "celltype.A") %>% pull(gene) %>% unique()
}

# Create a heatmap data matrix --------------------------------------------
htData <- Seurat::DoHeatmap(seuObj, features = topGenes, group.by = "cluster")$data %>% tidyr::drop_na()

{
    htData.mtx <- htData %>% 
        select(Feature, Cell, Expression) %>% 
        tidyr::pivot_wider(names_from = Cell, values_from = Expression, values_fill = NA) %>% 
        tibble::column_to_rownames("Feature")
    
    colAnnot <- htData %>% 
        select(Cell, Identity) %>% 
        distinct(Cell, .keep_all = TRUE) %>% 
        mutate(Identity = as.vector(Identity), Cell = as.vector(Cell)) %>% 
        filter(Cell %in% colnames(htData.mtx)) %>% 
        arrange(Identity) %>% 
        tibble::column_to_rownames("Cell")
    
    htData.mtx <- htData.mtx[, rownames(colAnnot)]
}


# Create and save the heatmap ---------------------------------------------
ht <- Heatmap(
    as.matrix(htData.mtx),# Convert the data matrix to a matrix format for the heatmap
    name = "Enhance Heatmap",# Name of the heatmap displayed in the legend
    show_column_names = FALSE,# Do not display the column names (e.g., sample names)
    row_names_side = 'left',# Display row names (e.g., gene names) on the left side
    column_split = as.factor(colAnnot$Identity),# Split columns based on the 'Identity' factor from colAnnot
    cluster_rows = TRUE,# Perform hierarchical clustering on the rows (genes)
    show_row_dend = FALSE,# Do not display the dendrogram for rows
    cluster_columns = TRUE,# Perform hierarchical clustering on the columns (samples)
    cluster_column_slices = FALSE,# Do not perform clustering on column slices after splitting
    show_column_dend = FALSE,# Do not display the dendrogram for columns
    show_parent_dend_line = FALSE,# Do not show the lines of the parent dendrogram
    top_annotation = HeatmapAnnotation(Idents = colAnnot$Identity, show_legend = FALSE),# Add a top annotation to the heatmap showing the identities, without a legend
    col = colorRampPalette(c("#ec00ec", "#010001", "#fbfb00"))(n = 100),# Define the color palette for the heatmap, ranging from magenta to black to yellow
    row_km = 5,# Split rows into 5 groups using k-means clustering
    row_names_gp = gpar(fontsize = 8),# Set the font size for the row names
    # Configure the heatmap legend with specific parameters
    heatmap_legend_param = list(
        at = c(-2, 0, 2),  # Legend ticks at -2, 0, and 2
        title = "Scaled.Expression",  # Title of the legend
        legend_height = unit(6, "cm"),  # Height of the legend
        title_position = "leftcenter-rot"  # Position of the title
    )
)

# Capture the heatmap 'ht' as a graphical object using 'grid.grabExpr' and store it in 'p'
p <- grid.grabExpr(draw(ht))

# Save the heatmap to a file
ggsave(filename = "heatmap_example.jpg", plot = p, width = 22, height = 12, dpi = 300)

Reference

DoHeatmap clustering specific genes and not top x genes · Issue #2261 · satijalab/seurat (github.com)