Command Line Tools¶

pyInfinityFlow¶

Usage¶

Simple¶

All Arguments¶

Here is an example of calling the command line program to run the pipeline specifying every argument. This is intended to be useful for quick copying and pasting of all command arguments to later remove selected and update values of modified arguments. The settings of the arguments here attempt to replicate how the original author’s processed their published¹ mouse lung dataset using the BioLegend Murine LEGENDScreen kit (Eg. use 50% of all measured events from InfinityMarker files and pool 10,000 cells from each InfinityMarker final into the final InfinityFlow object):

pyInfinityFlow --data_dir "pyInfinityFlow/data/fcs_files/" \
                --outdir "pyInfinityFlow/output_dir/" \
                --backbone_annotation "pyInfinityFlow/data/small_test_dataset_backbone_anno.csv" \
                --infinity_marker_annotation "pyInfinityFlow/data/small_test_dataset_infinity_marker_anno.csv" \
                --random_state 7 \
                --use_logicle_scaling True \
                --normalization_method "zscore" \
                --n_events_train 0 \
                --n_events_validate 0 \
                --ratio_for_validation 0.5 \
                --separate_backbone_reference None \
                --n_events_combine 10000 \
                --n_final 0 \
                --add_umap True \
                --find_clusters True \
                --find_markers True \
                --make_feature_plots True \
                --use_pca True \
                --n_pc 15 \
                --n_pc_plot_qc 50 \
                --save_h5ad True \
                --save_feather True \
                --save_file_handler True \
                --save_regression_models True \
                --verbosity 1 \
                --n_cores 12

Required Arguments¶

--data_dir (str): Directory in which the .fcs files are contained. Each InfinityMarker must be associated with an .fcs file in this directory. Each .fcs file must have the channel names present in the backbone_annotation file. It is expected that the values of the .fcs data are already compensated and don’t require adjustment with spillover. For example, the “export compensated values” feature from FlowJo could be used to export compensated Flow Cytometry data.

--outdir (str): Directory in which to save the outputs.

--backbone_annotation (str): Path to the backbone annotation file. It can be either a CSV (.csv) or TSV (.tsv or .txt) file with a header that annotates the column names. Each subsequent line annotates a backbone channel for the regression models, which will be used as predictors for the targets (InfinityMarkers). For each backbone channel, the following columns are annotated:

Reference_Backbone: the name of the backbone channel as it appears in the --separate_backbone_reference file (if a separate reference is used) or the InfinityMarker files if --n_events_combine is used to pool events from InfinityMarkers into a final InfinityFlow object
Query_Backbone: the name of the backbone channel as it is written in the InfinityMarker files (if the channel name of backbone parameters is different between InfinityMarker files, the pipeline must be run multiple times for each backbone channel layout)
Final_Name: the desired name parameter in the final InfinityFlow object

It is recommended to build this file in Excel and export as a .csv file. Here is an example of the structure of a “backbone_annotation” file:

Reference_Backbone	Query_Backbone	Final_Name
FJComp-APC-A	FJComp-APC-A	CD69-CD301b
FJComp-AlexaFluor700-A	FJComp-AlexaFluor700-A	MHCII
FJComp-BUV395-A	FJComp-BUV395-A	CD4
FJComp-BUV737-A	FJComp-BUV737-A	CD44
FJComp-BV421-A	FJComp-BV421-A	CD8
FJComp-BV510-A	FJComp-BV510-A	CD11c
FJComp-BV605-A	FJComp-BV605-A	CD11b
FJComp-BV650-A	FJComp-BV650-A	F480
FJComp-BV711-A	FJComp-BV711-A	Ly6C
FJComp-BV786-A	FJComp-BV786-A	Lineage
FJComp-GFP-A	FJComp-GFP-A	CD45a488
FJComp-PE-Cy7(yg)-A	FJComp-PE-Cy7(yg)-A	CD24
FJComp-PerCP-Cy5-5-A	FJComp-PerCP-Cy5-5-A	CD103

--infinity_marker_annotation (str): The path to the file which annotates each InfinityMarker. It can be either a CSV (.csv) or TSV (.tsv or .txt) file with a header that annotates the column names. Each subsequent line of the file annotates an InfinityMarker with the following columns:

File: the name of the file as it is saved in the --data_dir
Channel: the name of the channel, as it is written in the .fcs File
Name: the desired name to give to the channel in the final InfinityFlow object (should be unique)
Isotype: (OPTIONAL) the InfinityMarker that serves as the isotype for the given InfinityMarker (must match one of the other Name values in the third column)

It is recommended to build this file in Excel and export as a .csv file. Here is an example of the structure of an “infinity_marker_annotation” file:

File	Channel	Name	Isotype (Optional)
backbone_Plate1_Specimen_001_A11_A11_011_target_CD11b.fcs	FJComp-PE(yg)-A	CD11b	rIgG2b
backbone_Plate1_Specimen_001_A12_A12_012_target_CD11c.fcs	FJComp-PE(yg)-A	CD11c	AHIgG
backbone_Plate1_Specimen_001_B11_B11_023_target_CD27.fcs	FJComp-PE(yg)-A	CD27	AHIgG
backbone_Plate1_Specimen_001_B2_B02_014_target_CD16-32.fcs	FJComp-PE(yg)-A	CD16-32	rIgG2a
backbone_Plate1_Specimen_001_D2_D02_038_target_CD45R-B220.fcs	FJComp-PE(yg)-A	CD45R-B220	rIgG2a
backbone_Plate1_Specimen_001_E12_E12_060_target_CD71.fcs	FJComp-PE(yg)-A	CD71	rIgG2a
backbone_Plate1_Specimen_001_E2_E02_050_target_CD55.fcs	FJComp-PE(yg)-A	CD55	AHIgG
backbone_Plate1_Specimen_001_G10_G10_082_target_CD117 (c-kit).fcs	FJComp-PE(yg)-A	CD117	rIgG2b
backbone_Plate3_Specimen_001_B3_B03_015_target_Ly-6C.fcs	FJComp-PE(yg)-A	Ly-6C	rIgG2c
backbone_Plate3_Specimen_001_B5_B05_017_target_Ly-6G.fcs	FJComp-PE(yg)-A	Ly-6G	rIgG2a
backbone_Plate3_Specimen_001_F11_F11_071_target_Isotype_rIgG2a.fcs	FJComp-PE(yg)-A	rIgG2a	rIgG2a
backbone_Plate3_Specimen_001_F12_F12_072_target_Isotype_rIgG2b.fcs	FJComp-PE(yg)-A	rIgG2b	rIgG2b
backbone_Plate3_Specimen_001_G1_G01_073_target_Isotype_rIgG2c.fcs	FJComp-PE(yg)-A	rIgG2c	rIgG2c
backbone_Plate3_Specimen_001_F4_F04_064_target_Isotype_AHIgG.fcs	FJComp-PE(yg)-A	AHIgG	AHIgG

Optional Arguments¶

--random_state (int|None) (Default=None): Integer to specify the random_state of sampling, regression, and UMAP to make results more reproducible.

--use_logicle_scaling (True|False) (Default=True): Whether or not to apply logicle scaling to features that are typically fluorescence channels in Flow Cytometry and not common linear features (Eg. FSC-A, SSC-A, …)

--normalization_method (“zscore”|None) (Default="zscore"): Method used for normalizing backbone feature values before regression in an effort to reduce sample to sample batch effects.

--n_events_train (int) (Default=0): Integer to specify the number events to use for training. 0 is a special case in which all events from the file will be used, in which case, --n_events_validate must also be set to 0 and --ratio_for_validation must be greater than 0 and less than 1.0. The sum of --n_events_train and --n_events_validate must not exceed the number of events in any of the InfinityMarker .fcs files annotated in the --infinity_marker_annotation file.

--n_events_validate (int) (Default=0): Integer to specify the number events to use for validation. 0 is a special case in which all events from the file will be used, in which case, --n_events_train must also be set to 0 and --ratio_for_validation must be greater than 0 and less than 1.0. The sum of --n_events_train and --n_events_validate must not exceed the number of events in any of the InfinityMarker .fcs files annotated in the --infinity_marker_annotation file.

--ratio_for_validation (0 < float < 1) (Default=0.2): If --n_events_train and --n_events_validate are both 0, then all of the events from the fcs file will be used and this argument will will specify what percentage of the dataset should be used for validation and the remainder will be used for training.

--separate_backbone_reference (str | None) (Default=None): The .fcs file passed as a file path string. This can be used as an alternative to --n_events_combine. The regression will then be applied to the events in this file. Each of the Infinity Markers specified in the infinity_marker_annotation file and the original channel values for this --separate_backbone_reference will be in the final output.

--n_events_combine (int|None) (Default=None): As an alternative to using a separate, external, reference .fcs file, the --n_events_combine argument can be used to pool events from the InfinityMarker input .fcs files specified in the --infinity_marker_annotation file to sample events for the final InfinityFlow object. The --random_state argument will set the seed for this sampling. The resulting InfinityFlow object will be made up of an even sample of --n_events_combine from each unique InfinityMarker file. 0 is a special case in which all events from the InfinityMarker files will be pooled together.

--n_final (int) (Default=0): Specifies the number of events to include in the final InfinityFlow object. This will either sample from the --separate_backbone_reference file (if not set to None), in which case the value needs to be <= the number of events in that file; or from the pooled cells specified by --n_events_combine, in which case it needs to be less than the sum of the pooled events.

--add_umap (True|False) (Default=False): Boolean to specify if UMAP dimensionality reduction should be carried out on the final InfinityFlow object to reduce to 2 dimensions for visualization.

--find_clusters (True|False) (Default=False): Boolean to specify clustering should be done using Leiden clustering implemented through the Scanpy package.

--find_markers (True|False) (Default=False): Boolean to specify if MarkerFinder should be applied to find optimal markers for clusters. --find_clusters must be set to True to use this feature.

--make_feature_plots (True|False) (Default=False): Boolean to specify if each feature in the final InfinityFlow object should be plotted over the 2D UMAP embedding.

--use_pca (True|False) (Default=True): Boolean to specify if principal component ananlysis should be used to reduce the feature space prior to UMAP and clustering. This is suggested to save computation time.

--n_pc (int) (Default=15): Integer to specify the number principal components to use for UMAP and clustering. It is recommended to look at the PC-elbo curve in the outputs to refine the optimal number of principal components to use. This value must be less than the total number of features in the final InfinityFlow object (InfinityMarkers + backbone channels) defined in the --infinity_marker_annotation file and --backbone_annotation file.

--n_pc_plot_qc (int) (Default=50): Integer to specify the number principal components to plot in the elbo curve. Helpful for estimating the number of principal components to use downstream.

--save_h5ad (True|False) (Default=False): Boolean to specify if the final InfinityFlow object should be saved as an h5ad file. Useful for quick loading of the data into a Python anndata.AnnData object for downstream analyses with Scanpy.

--save_feather (True|False) (Default=False): Boolean to specify if the final InfinityFlow object should be saved as a DataFrame in a feather file. Useful for quick loading of the data into a Python Pandas DataFrame.

--save_file_handler (True|False) (Default=False): Boolean to specify whether or not to save the intermediate pyInfinityFlow.file_handler object, which stores data on how each of the InfinityMarker .fcs files were processed.

--save_regression_models (True|False) (Default=False): Boolean to specify whether or not to save the intermediate pyInfinityFlow.regression_models object, which stores the regression models and validation metrics for each InfinityMarker feature.

--verbosity (0|1|2|3) (Default=1): The level of verbosity with which to write to std-out. 0 = no print statements, to 3 = all debug print statements.

--n_cores (int) (Default=1): The number of cores to use, which can increase the speed of regression fitting, UMAP dimensionality reduction, and Leiden clustering.

Outputs¶

The outputs of the pipeline are written to the path specified by the --outdir argument in the following tree structure: