bio-flow-cytometry-cytometry-qc

name	bio-flow-cytometry-cytometry-qc
description	Comprehensive quality control for flow cytometry and CyTOF data. Covers flow rate stability, signal drift, margin events, dead cell exclusion, and batch QC. Use when assessing acquisition quality or identifying problematic samples before analysis.
tool_type	r
primary_tool	flowAI

Cytometry QC

Automated QC with flowAI

library(flowAI)
library(flowCore)

# Load FCS file
ff <- read.FCS('sample.fcs')

# Run automated QC
# Checks: flow rate, signal stability, dynamic range
qc_result <- flow_auto_qc(
    ff,
    folder_results = 'qc_output/',
    fcs_QC = TRUE,       # Export QC'd FCS
    html_report = TRUE,  # Generate HTML report
    mini_report = TRUE   # Also make summary
)

# Get cleaned data
ff_clean <- qc_result$fcs

# QC metrics
cat('Original events:', nrow(ff), '\n')
cat('After QC:', nrow(ff_clean), '\n')
cat('Removed:', nrow(ff) - nrow(ff_clean), '(',
    round((1 - nrow(ff_clean)/nrow(ff)) * 100, 1), '%)\n')

Flow Rate Stability

# Check for acquisition issues via flow rate
check_flow_rate <- function(ff, time_channel = 'Time') {
    expr <- exprs(ff)
    time <- expr[, time_channel]

    # Bin by time
    n_bins <- 50
    bins <- cut(time, breaks = n_bins, labels = FALSE)

    # Events per bin (flow rate proxy)
    events_per_bin <- table(bins)
    flow_rate <- as.numeric(events_per_bin)

    # Calculate metrics
    cv <- sd(flow_rate) / mean(flow_rate) * 100

    # Detect anomalies (>2 SD from mean)
    z_scores <- abs(scale(flow_rate))
    anomalies <- which(z_scores > 

    
        mean_rate  meanflow_rate
        cv_percent  cv
        anomaly_bins  anomalies
        stable  cv    anomalies  
    


flow_qc  check_flow_rateff
cat flow_qccv_percent  
cat flow_qcstable 

Signal Drift Detection

# Detect signal drift over acquisition time
detect_signal_drift <- function(ff, channels, time_channel = 'Time') {
    expr <- exprs(ff)
    time <- expr[, time_channel]

    n_bins <- 20
    bins <- cut(time, breaks = n_bins, labels = FALSE)

    drift_results <- lapply(channels, function(ch) {
        bin_medians <- tapply(expr[, ch], bins, median, na.rm = TRUE)

        # Linear trend
        trend <- lm(bin_medians ~ seq_along(bin_medians))
        slope <- coef(trend
        r_squared  summarytrendr.squared

        
        pct_change  tailbin_medians   headbin_medians   headbin_medians   

        
            channel  ch
            slope  slope
            r_squared  r_squared
            percent_change  pct_change
            drift_detected  pct_change    r_squared  
        
    

    drift_results  channels
    drift_results


marker_channels     
drift  detect_signal_driftff marker_channels
 ch  drift 
     driftchdrift_detected 
        cat ch  driftchpercent_change  
    

Margin Events Removal

# Remove events at detector saturation limits
remove_margin_events <- function(ff, channels = NULL) {
    expr <- exprs(ff)

    if (is.null(channels)) {
        channels <- colnames(expr)
    }

    # Get channel ranges from FCS parameters
    params <- parameters(ff)

    margin_mask <- rep(FALSE, nrow(expr))

    for (ch in channels) {
        if (ch %in% colnames(expr)) {
            # Get max range from parameters
            idx <- match(ch, params@data$name)
             idx 
                max_val  paramsdataidx
                
                margin_mask  margin_mask  expr ch  max_val    expr ch  
            
        
    

    cat margin_mask  meanmargin_mask    
    ffmargin_mask 


ff_no_margin  remove_margin_eventsff  

Dead Cell Exclusion

# Exclude dead cells using viability marker
exclude_dead_cells <- function(ff, viability_channel, threshold = NULL) {
    expr <- exprs(ff)
    viability <- expr[, viability_channel]

    if (is.null(threshold)) {
        # Auto-threshold using bimodal distribution
        # Dead cells have higher viability dye uptake
        threshold <- quantile(viability, 0.9)
    }

    live_mask <- viability < threshold

    cat('Total events:', length(live_mask), '\n')
    cat('Live cells:', sum(live_mask), '(', round(meanlive_mask    
    cat live_mask  meanlive_mask    

    fflive_mask 



ff_live  exclude_dead_cellsff 

CyTOF-Specific QC

# CyTOF-specific quality metrics
cytof_qc <- function(ff) {
    expr <- exprs(ff)

    # Event length check (cell size proxy)
    if ('Event_length' %in% colnames(expr)) {
        event_length <- expr[, 'Event_length']

        # Typical single cells: 15-45
        good_length <- event_length >= 15 & event_length <= 45
        cat('Event length filter:', sum(good_length), '/', length(good_length),
            '(', round(mean(good_length) * 100, 1), 
    

    
    dna_channels  grep colnamesexpr value  
     dna_channels   
        dna_signal  rowMeansexpr dna_channels drop  

        
        has_dna  dna_signal  quantiledna_signal 
        cat has_dna  meanhas_dna    
    

    
    gauss_channels  grep colnamesexpr value  
     gauss_channels   
        cat
    


cytof_qcff

Batch QC Summary

library(dplyr)

# Generate QC summary for batch of files
batch_qc_summary <- function(fcs_files) {
    results <- lapply(fcs_files, function(f) {
        ff <- read.FCS(f)

        # Basic metrics
        n_events <- nrow(ff)

        # Flow rate CV
        flow_qc <- check_flow_rate(ff)

        # Signal range check
        expr <- exprs(ff)
        signal_channels <- grep('(FSC|SSC)', colnames(expr), value = TRUE, invert = TRUE)
        median_signals <- apply(expr[, signal_channels, drop = FALSE], 2 median

        data.frame
            file  basenamef
            events  n_events
            flow_rate_cv  flow_qccv_percent
            flow_stable  flow_qcstable
            median_signal  meanmedian_signals na.rm  
        
    

    summary_df  do.callrbind results

    
    summary_dfoutlier  withsummary_df
        events  medianevents   
        events  medianevents   
        flow_rate_cv  
    

    summary_df



fcs_files  list.files pattern   full.names  
qc_summary  batch_qc_summaryfcs_files
printqc_summary


cat
printqc_summaryqc_summaryoutlier 

Visualization

library(ggplot2)

# Flow rate plot
plot_flow_rate <- function(ff, time_channel = 'Time') {
    expr <- exprs(ff)
    time <- expr[, time_channel]

    n_bins <- 100
    bins <- cut(time, breaks = n_bins, labels = FALSE)
    events_per_bin <- table(bins)

    plot_data <- data.frame(
        bin = as.numeric(names(events_per_bin)),
        events = as.numeric(events_per_bin)
    )

    ggplot(plot_data, aes(x = bin, y = events)) +
        geom_line 
        geom_smoothmethod   color   se   
        theme_bw 
        labstitle   x   y  



plot_signal_stability  ff channel time_channel   
    expr  exprsff

    n_bins  50
    bins  cutexpr time_channel breaks  n_bins labels  
    bin_stats  tapplyexpr channel bins x 
        median  medianx q25  quantilex  q75  quantilex 
    

    plot_data  data.frame
        bin  bin_stats
        median  sapplybin_stats  
        q25  sapplybin_stats  
        q75  sapplybin_stats  
    

    ggplotplot_data aesx  bin 
        geom_ribbonaesymin  q25 ymax  q75 alpha   
        geom_lineaesy  median color   
        theme_bw 
        labstitle  paste channel x   y  



p1  plot_flow_rateff
ggsave p1 width   height  

p2  plot_signal_stabilityff 
ggsave p2 width   height  

QC Report Generation

# Generate comprehensive QC report
generate_qc_report <- function(ff, output_file = 'qc_report.txt') {
    sink(output_file)

    cat('=== FLOW CYTOMETRY QC REPORT ===\n\n')
    cat('File:', description(ff)$`$FIL`, '\n')
    cat('Date:', description(ff)$`$DATE`, '\n')
    cat('Total events:', nrow(ff), '\n\n')

    cat('--- Flow Rate ---\n')
    flow_qc <- check_flow_rate(ff)
    cat('CV:', round(flow_qc$cv_percent, 1), 
    cat ifelseflow_qcstable   

    cat
    expr  exprsff
     ch  colnamesexpr ncolexpr 
        catch  medianexpr ch  
    

    sink
    cat output_file 


generate_qc_reportff

Related Skills

Workflow order: cytometry-qc → doublet-detection → bead-normalization → clustering

• compensation-transformation - Data preprocessing before QC
• doublet-detection - Run after QC: remove doublet events
• bead-normalization - Run after doublet removal: correct signal drift
• clustering-phenotyping - Analysis after all preprocessing