flowio
À propos
FlowIO analyse les fichiers FCS (versions 2.0 à 3.1) pour extraire les données d'événements de cytométrie en flux dans des tableaux NumPy ainsi que les métadonnées. Il permet la conversion vers des formats tels que CSV ou DataFrames pour les pipelines de prétraitement des données. Utilisez cette compétence pour les services backend nécessitant de lire, écrire ou convertir le contenu des fichiers FCS.
Installation rapide
Claude Code
Recommandénpx skills add K-Dense-AI/claude-scientific-skills -a claude-code/plugin add https://github.com/K-Dense-AI/claude-scientific-skillsgit clone https://github.com/K-Dense-AI/claude-scientific-skills.git ~/.claude/skills/flowioCopiez et collez cette commande dans Claude Code pour installer cette compétence
Documentation
FlowIO: Flow Cytometry Standard File Handler
Overview
FlowIO is a lightweight Python library for reading and writing Flow Cytometry Standard (FCS) files. Parse FCS metadata, extract event data, and create new FCS files with minimal dependencies. The library supports FCS versions 2.0, 3.0, and 3.1, making it ideal for backend services, data pipelines, and basic cytometry file operations.
When to Use This Skill
This skill should be used when:
- FCS files requiring parsing or metadata extraction
- Flow cytometry data needing conversion to NumPy arrays
- Event data requiring export to FCS format
- Multi-dataset FCS files needing separation
- Channel information extraction (scatter, fluorescence, time)
- Cytometry file validation or inspection
- Pre-processing workflows before advanced analysis
Related Tools: For advanced flow cytometry analysis including compensation, gating, and FlowJo/GatingML support, recommend FlowKit library as a companion to FlowIO.
Installation
uv pip install flowio
Requires Python 3.9 or later.
Quick Start
Basic File Reading
from flowio import FlowData
# Read FCS file
flow_data = FlowData('experiment.fcs')
# Access basic information
print(f"FCS Version: {flow_data.version}")
print(f"Events: {flow_data.event_count}")
print(f"Channels: {flow_data.pnn_labels}")
# Get event data as NumPy array
events = flow_data.as_array() # Shape: (events, channels)
Creating FCS Files
import numpy as np
from flowio import create_fcs
# Prepare data
data = np.array([[100, 200, 50], [150, 180, 60]]) # 2 events, 3 channels
channels = ['FSC-A', 'SSC-A', 'FL1-A']
# Create FCS file
create_fcs('output.fcs', data, channels)
Core Workflows
Reading and Parsing FCS Files
The FlowData class provides the primary interface for reading FCS files.
Standard Reading:
from flowio import FlowData
# Basic reading
flow = FlowData('sample.fcs')
# Access attributes
version = flow.version # '3.0', '3.1', etc.
event_count = flow.event_count # Number of events
channel_count = flow.channel_count # Number of channels
pnn_labels = flow.pnn_labels # Short channel names
pns_labels = flow.pns_labels # Descriptive stain names
# Get event data
events = flow.as_array() # Preprocessed (gain, log scaling applied)
raw_events = flow.as_array(preprocess=False) # Raw data
Memory-Efficient Metadata Reading:
When only metadata is needed (no event data):
# Only parse TEXT segment, skip DATA and ANALYSIS
flow = FlowData('sample.fcs', only_text=True)
# Access metadata
metadata = flow.text # Dictionary of TEXT segment keywords
print(metadata.get('$DATE')) # Acquisition date
print(metadata.get('$CYT')) # Instrument name
Handling Problematic Files:
Some FCS files have offset discrepancies or errors:
# Ignore offset discrepancies between HEADER and TEXT sections
flow = FlowData('problematic.fcs', ignore_offset_discrepancy=True)
# Use HEADER offsets instead of TEXT offsets
flow = FlowData('problematic.fcs', use_header_offsets=True)
# Ignore offset errors entirely
flow = FlowData('problematic.fcs', ignore_offset_error=True)
Excluding Null Channels:
# Exclude specific channels during parsing
flow = FlowData('sample.fcs', null_channel_list=['Time', 'Null'])
Extracting Metadata and Channel Information
FCS files contain rich metadata in the TEXT segment.
Common Metadata Keywords:
flow = FlowData('sample.fcs')
# File-level metadata
text_dict = flow.text
acquisition_date = text_dict.get('$DATE', 'Unknown')
instrument = text_dict.get('$CYT', 'Unknown')
data_type = flow.data_type # 'I', 'F', 'D', 'A'
# Channel metadata
for i in range(flow.channel_count):
pnn = flow.pnn_labels[i] # Short name (e.g., 'FSC-A')
pns = flow.pns_labels[i] # Descriptive name (e.g., 'Forward Scatter')
pnr = flow.pnr_values[i] # Range/max value
print(f"Channel {i}: {pnn} ({pns}), Range: {pnr}")
Channel Type Identification:
FlowIO automatically categorizes channels:
# Get indices by channel type
scatter_idx = flow.scatter_indices # [0, 1] for FSC, SSC
fluoro_idx = flow.fluoro_indices # [2, 3, 4] for FL channels
time_idx = flow.time_index # Index of time channel (or None)
# Access specific channel types
events = flow.as_array()
scatter_data = events[:, scatter_idx]
fluorescence_data = events[:, fluoro_idx]
ANALYSIS Segment:
If present, access processed results:
if flow.analysis:
analysis_keywords = flow.analysis # Dictionary of ANALYSIS keywords
print(analysis_keywords)
Creating New FCS Files
Generate FCS files from NumPy arrays or other data sources.
Basic Creation:
import numpy as np
from flowio import create_fcs
# Create event data (rows=events, columns=channels)
events = np.random.rand(10000, 5) * 1000
# Define channel names
channel_names = ['FSC-A', 'SSC-A', 'FL1-A', 'FL2-A', 'Time']
# Create FCS file
create_fcs('output.fcs', events, channel_names)
With Descriptive Channel Names:
# Add optional descriptive names (PnS)
channel_names = ['FSC-A', 'SSC-A', 'FL1-A', 'FL2-A', 'Time']
descriptive_names = ['Forward Scatter', 'Side Scatter', 'FITC', 'PE', 'Time']
create_fcs('output.fcs',
events,
channel_names,
opt_channel_names=descriptive_names)
With Custom Metadata:
# Add TEXT segment metadata
metadata = {
'$SRC': 'Python script',
'$DATE': '19-OCT-2025',
'$CYT': 'Synthetic Instrument',
'$INST': 'Laboratory A'
}
create_fcs('output.fcs',
events,
channel_names,
opt_channel_names=descriptive_names,
metadata=metadata)
Note: FlowIO exports as FCS 3.1 with single-precision floating-point data.
Exporting Modified Data
Modify existing FCS files and re-export them.
Approach 1: Using write_fcs() Method:
from flowio import FlowData
# Read original file
flow = FlowData('original.fcs')
# Write with updated metadata
flow.write_fcs('modified.fcs', metadata={'$SRC': 'Modified data'})
Approach 2: Extract, Modify, and Recreate:
For modifying event data:
from flowio import FlowData, create_fcs
# Read and extract data
flow = FlowData('original.fcs')
events = flow.as_array(preprocess=False)
# Modify event data
events[:, 0] = events[:, 0] * 1.5 # Scale first channel
# Create new FCS file with modified data
create_fcs('modified.fcs',
events,
flow.pnn_labels,
opt_channel_names=flow.pns_labels,
metadata=flow.text)
Handling Multi-Dataset FCS Files
Some FCS files contain multiple datasets in a single file.
Detecting Multi-Dataset Files:
from flowio import FlowData, MultipleDataSetsError
try:
flow = FlowData('sample.fcs')
except MultipleDataSetsError:
print("File contains multiple datasets")
# Use read_multiple_data_sets() instead
Reading All Datasets:
from flowio import read_multiple_data_sets
# Read all datasets from file
datasets = read_multiple_data_sets('multi_dataset.fcs')
print(f"Found {len(datasets)} datasets")
# Process each dataset
for i, dataset in enumerate(datasets):
print(f"\nDataset {i}:")
print(f" Events: {dataset.event_count}")
print(f" Channels: {dataset.pnn_labels}")
# Get event data for this dataset
events = dataset.as_array()
print(f" Shape: {events.shape}")
print(f" Mean values: {events.mean(axis=0)}")
Reading Specific Dataset:
from flowio import FlowData
# Read first dataset (nextdata_offset=0)
first_dataset = FlowData('multi.fcs', nextdata_offset=0)
# Read second dataset using NEXTDATA offset from first
next_offset = int(first_dataset.text['$NEXTDATA'])
if next_offset > 0:
second_dataset = FlowData('multi.fcs', nextdata_offset=next_offset)
Data Preprocessing
FlowIO applies standard FCS preprocessing transformations when preprocess=True.
Preprocessing Steps:
- Gain Scaling: Multiply values by PnG (gain) keyword
- Logarithmic Transformation: Apply PnE exponential transformation if present
- Formula:
value = a * 10^(b * raw_value)where PnE = "a,b"
- Formula:
- Time Scaling: Convert time values to appropriate units
Controlling Preprocessing:
# Preprocessed data (default)
preprocessed = flow.as_array(preprocess=True)
# Raw data (no transformations)
raw = flow.as_array(preprocess=False)
Error Handling
Handle common FlowIO exceptions appropriately.
from flowio import (
FlowData,
FCSParsingError,
DataOffsetDiscrepancyError,
MultipleDataSetsError
)
try:
flow = FlowData('sample.fcs')
events = flow.as_array()
except FCSParsingError as e:
print(f"Failed to parse FCS file: {e}")
# Try with relaxed parsing
flow = FlowData('sample.fcs', ignore_offset_error=True)
except DataOffsetDiscrepancyError as e:
print(f"Offset discrepancy detected: {e}")
# Use ignore_offset_discrepancy parameter
flow = FlowData('sample.fcs', ignore_offset_discrepancy=True)
except MultipleDataSetsError as e:
print(f"Multiple datasets detected: {e}")
# Use read_multiple_data_sets instead
from flowio import read_multiple_data_sets
datasets = read_multiple_data_sets('sample.fcs')
except Exception as e:
print(f"Unexpected error: {e}")
Common Use Cases
Inspecting FCS File Contents
Quick exploration of FCS file structure:
from flowio import FlowData
flow = FlowData('unknown.fcs')
print("=" * 50)
print(f"File: {flow.name}")
print(f"Version: {flow.version}")
print(f"Size: {flow.file_size:,} bytes")
print("=" * 50)
print(f"\nEvents: {flow.event_count:,}")
print(f"Channels: {flow.channel_count}")
print("\nChannel Information:")
for i, (pnn, pns) in enumerate(zip(flow.pnn_labels, flow.pns_labels)):
ch_type = "scatter" if i in flow.scatter_indices else \
"fluoro" if i in flow.fluoro_indices else \
"time" if i == flow.time_index else "other"
print(f" [{i}] {pnn:10s} | {pns:30s} | {ch_type}")
print("\nKey Metadata:")
for key in ['$DATE', '$BTIM', '$ETIM', '$CYT', '$INST', '$SRC']:
value = flow.text.get(key, 'N/A')
print(f" {key:15s}: {value}")
Batch Processing Multiple Files
Process a directory of FCS files:
from pathlib import Path
from flowio import FlowData
import pandas as pd
# Find all FCS files
fcs_files = list(Path('data/').glob('*.fcs'))
# Extract summary information
summaries = []
for fcs_path in fcs_files:
try:
flow = FlowData(str(fcs_path), only_text=True)
summaries.append({
'filename': fcs_path.name,
'version': flow.version,
'events': flow.event_count,
'channels': flow.channel_count,
'date': flow.text.get('$DATE', 'N/A')
})
except Exception as e:
print(f"Error processing {fcs_path.name}: {e}")
# Create summary DataFrame
df = pd.DataFrame(summaries)
print(df)
Converting FCS to CSV
Export event data to CSV format:
from flowio import FlowData
import pandas as pd
# Read FCS file
flow = FlowData('sample.fcs')
# Convert to DataFrame
df = pd.DataFrame(
flow.as_array(),
columns=flow.pnn_labels
)
# Add metadata as attributes
df.attrs['fcs_version'] = flow.version
df.attrs['instrument'] = flow.text.get('$CYT', 'Unknown')
# Export to CSV
df.to_csv('output.csv', index=False)
print(f"Exported {len(df)} events to CSV")
Filtering Events and Re-exporting
Apply filters and save filtered data:
from flowio import FlowData, create_fcs
import numpy as np
# Read original file
flow = FlowData('sample.fcs')
events = flow.as_array(preprocess=False)
# Apply filtering (example: threshold on first channel)
fsc_idx = 0
threshold = 500
mask = events[:, fsc_idx] > threshold
filtered_events = events[mask]
print(f"Original events: {len(events)}")
print(f"Filtered events: {len(filtered_events)}")
# Create new FCS file with filtered data
create_fcs('filtered.fcs',
filtered_events,
flow.pnn_labels,
opt_channel_names=flow.pns_labels,
metadata={**flow.text, '$SRC': 'Filtered data'})
Extracting Specific Channels
Extract and process specific channels:
from flowio import FlowData
import numpy as np
flow = FlowData('sample.fcs')
events = flow.as_array()
# Extract fluorescence channels only
fluoro_indices = flow.fluoro_indices
fluoro_data = events[:, fluoro_indices]
fluoro_names = [flow.pnn_labels[i] for i in fluoro_indices]
print(f"Fluorescence channels: {fluoro_names}")
print(f"Shape: {fluoro_data.shape}")
# Calculate statistics per channel
for i, name in enumerate(fluoro_names):
channel_data = fluoro_data[:, i]
print(f"\n{name}:")
print(f" Mean: {channel_data.mean():.2f}")
print(f" Median: {np.median(channel_data):.2f}")
print(f" Std Dev: {channel_data.std():.2f}")
Best Practices
- Memory Efficiency: Use
only_text=Truewhen event data is not needed - Error Handling: Wrap file operations in try-except blocks for robust code
- Multi-Dataset Detection: Check for MultipleDataSetsError and use appropriate function
- Preprocessing Control: Explicitly set
preprocessparameter based on analysis needs - Offset Issues: If parsing fails, try
ignore_offset_discrepancy=Trueparameter - Channel Validation: Verify channel counts and names match expectations before processing
- Metadata Preservation: When modifying files, preserve original TEXT segment keywords
Advanced Topics
Understanding FCS File Structure
FCS files consist of four segments:
- HEADER: FCS version and byte offsets for other segments
- TEXT: Key-value metadata pairs (delimiter-separated)
- DATA: Raw event data (binary/float/ASCII format)
- ANALYSIS (optional): Results from data processing
Access these segments via FlowData attributes:
flow.header- HEADER segmentflow.text- TEXT segment keywordsflow.events- DATA segment (as bytes)flow.analysis- ANALYSIS segment keywords (if present)
Detailed API Reference
For comprehensive API documentation including all parameters, methods, exceptions, and FCS keyword reference, consult the detailed reference file:
Read: references/api_reference.md
The reference includes:
- Complete FlowData class documentation
- All utility functions (read_multiple_data_sets, create_fcs)
- Exception classes and handling
- FCS file structure details
- Common TEXT segment keywords
- Extended example workflows
When working with complex FCS operations or encountering unusual file formats, load this reference for detailed guidance.
Integration Notes
NumPy Arrays: All event data is returned as NumPy ndarrays with shape (events, channels)
Pandas DataFrames: Easily convert to DataFrames for analysis:
import pandas as pd
df = pd.DataFrame(flow.as_array(), columns=flow.pnn_labels)
FlowKit Integration: For advanced analysis (compensation, gating, FlowJo support), use FlowKit library which builds on FlowIO's parsing capabilities
Web Applications: FlowIO's minimal dependencies make it ideal for web backend services processing FCS uploads
Troubleshooting
Problem: "Offset discrepancy error"
Solution: Use ignore_offset_discrepancy=True parameter
Problem: "Multiple datasets error"
Solution: Use read_multiple_data_sets() function instead of FlowData constructor
Problem: Out of memory with large files
Solution: Use only_text=True for metadata-only operations, or process events in chunks
Problem: Unexpected channel counts
Solution: Check for null channels; use null_channel_list parameter to exclude them
Problem: Cannot modify event data in place
Solution: FlowIO doesn't support direct modification; extract data, modify, then use create_fcs() to save
Summary
FlowIO provides essential FCS file handling capabilities for flow cytometry workflows. Use it for parsing, metadata extraction, and file creation. For simple file operations and data extraction, FlowIO is sufficient. For complex analysis including compensation and gating, integrate with FlowKit or other specialized tools.
Dépôt GitHub
Compétences associées
llamaguard
AutreLlamaGuard est le modèle de Meta, doté de 7 à 8 milliards de paramètres, conçu pour modérer les entrées et sorties des LLM selon six catégories de sécurité comme la violence et les discours haineux. Il offre une précision de 94 à 95 % et peut être déployé avec vLLM, Hugging Face ou Amazon SageMaker. Utilisez cette compétence pour intégrer facilement le filtrage de contenu et des garde-fous de sécurité dans vos applications d'IA.
cost-optimization
AutreCette compétence de Claude aide les développeurs à optimiser les coûts du cloud grâce au redimensionnement des ressources, aux stratégies d'étiquetage et à l'analyse des dépenses. Elle fournit un cadre pour réduire les dépenses cloud et mettre en œuvre une gouvernance des coûts sur AWS, Azure et GCP. Utilisez-la lorsque vous devez analyser les coûts d'infrastructure, redimensionner les ressources ou respecter des contraintes budgétaires.
quantizing-models-bitsandbytes
AutreCette compétence quantifie les LLMs en précision 8 bits ou 4 bits à l'aide de bitsandbytes, permettant une réduction de 50 à 75 % de la mémoire utilisée avec une perte de précision minime. Elle est idéale pour exécuter des modèles plus volumineux sur une mémoire GPU limitée ou pour accélérer l'inférence, prenant en charge des formats comme INT8, NF4 et FP4. La compétence s'intègre à HuggingFace Transformers et permet l'entraînement QLoRA ainsi que l'utilisation d'optimiseurs en 8 bits.
dispatching-parallel-agents
AutreCette compétence Claude déploie plusieurs agents pour enquêter et résoudre simultanément 3 problèmes indépendants ou plus. Elle est conçue pour des scénarios impliquant des défaillances non liées qui peuvent être résolues sans état partagé ni dépendances. La capacité fondamentale est la résolution de problèmes en parallèle, en assignant un agent par domaine problématique indépendant afin de maximiser l'efficacité.
