/flowio-flow-cytometry

name: flowio-flow-cytometry
description: "Parse/write FCS (Flow Cytometry) files v2.0-3.1. Events as NumPy, channel metadata, multi-dataset files, CSV/FCS export. Use FlowKit for gating/compensation."
license: BSD-3-Clause

FlowIO — Flow Cytometry File Handler

Overview

FlowIO is a lightweight Python library for reading and writing Flow Cytometry Standard (FCS) files. It parses FCS metadata, extracts event data as NumPy arrays, and creates new FCS files. Supports FCS versions 2.0, 3.0, and 3.1. Minimal dependencies — ideal for data pipelines and preprocessing before advanced analysis.

When to Use

• Parsing FCS files to extract event data as NumPy arrays
• Reading channel metadata (names, ranges, types) from FCS files
• Converting flow cytometry data to pandas DataFrames or CSV
• Creating new FCS files from NumPy arrays or processed data
• Handling multi-dataset FCS files (separating combined datasets)
• Batch processing directories of FCS files
• Preprocessing flow cytometry data before downstream analysis
• For **compensation, gating, and FlowJo workspace support**, use FlowKit instead
• For **advanced cytometry visualization** (density plots, gating plots), use matplotlib or plotly

Prerequisites

pip install flowio numpy pandas

Requires Python 3.9+. No compiled dependencies — installs on any platform.

Quick Start

from flowio import FlowData

flow = FlowData("experiment.fcs")
print(f"Events: {flow.event_count}, Channels: {flow.channel_count}")
print(f"Channels: {flow.pnn_labels}")

events = flow.as_array()  # Shape: (n_events, n_channels)
print(f"Data shape: {events.shape}")

Core API

1. Reading FCS Files

The `FlowData` class is the primary interface for reading FCS files.

from flowio import FlowData

# Standard reading
flow = FlowData("sample.fcs")
print(f"Version: {flow.version}")          # '3.0', '3.1', etc.
print(f"Events: {flow.event_count}")
print(f"Channels: {flow.channel_count}")

# Event data
events = flow.as_array()                   # Preprocessed (gain, log scaling)
raw = flow.as_array(preprocess=False)      # Raw values
print(f"Shape: {events.shape}")            # (n_events, n_channels)

# Memory-efficient: metadata only (skip DATA segment)
flow_meta = FlowData("sample.fcs", only_text=True)
print(f"Instrument: {flow_meta.text.get('$CYT', 'Unknown')}")

# Handle problematic files
flow = FlowData("bad.fcs", ignore_offset_discrepancy=True)
flow = FlowData("bad.fcs", use_header_offsets=True)

# Exclude null channels
flow = FlowData("sample.fcs", null_channel_list=["Time", "Null"])

2. Channel Metadata

Extract channel names, types, and ranges from FCS files.

flow = FlowData("sample.fcs")

# Channel names
pnn = flow.pnn_labels   # Short names: ['FSC-A', 'SSC-A', 'FL1-A', ...]
pns = flow.pns_labels   # Descriptive: ['Forward Scatter', 'Side Scatter', 'FITC', ...]
pnr = flow.pnr_values   # Range/max values per channel

# Channel type indices
scatter_idx = flow.scatter_indices   # [0, 1] — FSC, SSC
fluoro_idx = flow.fluoro_indices     # [2, 3, 4] — fluorescence channels
time_idx = flow.time_index           # Time channel index (or None)

# Access by type
events = flow.as_array()
scatter_data = events[:, scatter_idx]
fluoro_data = events[:, fluoro_idx]

# Full metadata (TEXT segment dictionary)
text = flow.text
print(f"Date: {text.get('$DATE', 'N/A')}")
print(f"Instrument: {text.get('$CYT', 'N/A')}")

3. Creating FCS Files

Generate new FCS files from NumPy arrays.

import numpy as np
from flowio import create_fcs

# Basic creation
events = np.random.rand(10000, 5) * 1000
channels = ["FSC-A", "SSC-A", "FL1-A", "FL2-A", "Time"]
create_fcs("output.fcs", events, channels)

# With descriptive names and metadata
create_fcs(
    "output.fcs",
    events,
    channels,
    opt_channel_names=["Forward Scatter", "Side Scatter", "FITC", "PE", "Time"],
    metadata={"$SRC": "Python pipeline", "$DATE": "17-FEB-2026", "$CYT": "Synthetic"},
)
# Output: FCS 3.1, single-precision float

4. Multi-Dataset FCS Files

Handle FCS files containing multiple datasets.

from flowio import FlowData, read_multiple_data_sets, MultipleDataSetsError

# Detect multi-dataset files
try:
    flow = FlowData("sample.fcs")
except MultipleDataSetsError:
    datasets = read_multiple_data_sets("sample.fcs")
    print(f"Found {len(datasets)} datasets")
    for i, ds in enumerate(datasets):
        print(f"Dataset {i}: {ds.event_count} events, {ds.channel_count} channels")
        events = ds.as_array()

# Read specific dataset by offset
first = FlowData("multi.fcs", nextdata_offset=0)
next_offset = int(first.text.get("$NEXTDATA", "0"))
if next_offset > 0:
    second = FlowData("multi.fcs", nextdata_offset=next_offset)

5. Modifying and Re-Exporting

Read, modify, and save FCS data.

from flowio import FlowData, create_fcs

# Read original
flow = FlowData("original.fcs")
events = flow.as_array(preprocess=False)  # Use raw for modification

# Filter events (e.g., threshold on FSC)
mask = events[:, 0] > 500
filtered = events[mask]
print(f"Before: {len(events)}, After: {len(filtered)}")

# Save filtered data as new FCS
create_fcs(
    "filtered.fcs",
    filtered,
    flow.pnn_labels,
    opt_channel_names=flow.pns_labels,
    metadata={**flow.text, "$SRC": "Filtered"},
)

# Or write with updated metadata (no event modification)
flow.write_fcs("updated.fcs", metadata={"$SRC": "Updated"})

Key Concepts

FCS File Structure

FCS files consist of four segments:

| Segment | Content | FlowData attribute | |---------|---------|-------------------| | HEADER | Version, byte offsets | `flow.header` | | TEXT | Key-value metadata (`$DATE`, `$CYT`, channel names) | `flow.text` | | DATA | Event data (binary/float) | `flow.events` (bytes), `flow.as_array()` | | ANALYSIS | Optional processed results | `flow.analysis` |

Preprocessing (as_array)

When `preprocess=True` (default),