GitHub - eiriniar/CellCnn at python3 (original) (raw)

CellCnn

Installation

This branch contains a Python 3.7 implementation of CellCnn. We recommend using it withpipenv, by following the steps below:

1. Clone the CellCnn repository and checkout the python3 branch:
   git clone -b python3 https://github.com/eiriniar/CellCnn.git
2. Go to the CellCnn root directory:
   cd CellCnn
3. Install CellCnn and its dependencies:
   pipenv install

The above steps have to be performed only once. Then, each time you want to perform a CellCnn analysis, go to the CellCnn root directory and activate the pipenv virtual environment by running:

1. pipenv shell

Usage

Examples are provided in the subfolder CellCnn/cellCnn/examples.

Alternatively, for the analysis of mass/flow cytometry samples, CellCnn can be run from the command line. To get a list of command line options please run:

python run_analysis.py --help

For a CellCnn analysis with default settings only two arguments have to be provided:

python run_analysis.py -f fcs_samples_with_labels.csv -m markers.csv

The first input argument is a two-column CSV file, where the first column specifies input sample filenames and the second column the corresponding class labels. An example file is provided in CellCnn/cellCnn/examples/NK_fcs_samples_with_labels.csv.

The second input argument is a CSV file containing the names of markers/channels that should be used for the analysis. An example file is provided in CellCnn/cellCnn/examples/NK_markers.csv.

For example, to perform the analysis outlined in CellCnn/cellCnn/examples/NK_cell.ipynbfrom the command line, you can run the following (assuming your current directory is CellCnn/cellCnn/examples):

python ../run_analysis.py -f NK_fcs_samples_with_labels.csv -m NK_markers.csv -i NK_cell_dataset/gated_NK/ -o outdir_NK --export_csv --group_a CMV- --group_b CMV+ --verbose 0

The above command performs a binary classification CellCnn analysis, exports the learned filter weights as CSV files in the directory outdir_NK/exported_filter_weights and generates result plots in outdir_NK/plots. The following plots are generated:

filter_plots

• clustered_filter_weights.pdf :
  Filter weight vectors from all trained networks that pass a validation accuracy threshold, grouped in clusters via hierarchical clustering. Each row corresponds to a filter. The last column(s) indicate the weight(s) connecting each filter to the output class(es). Indices on the y-axis indicate the filter cluster memberships, as a result of the hierarchical clustering procedure.
• consensus_filter_weights.pdf :
  One representative filter per cluster is chosen (the filter with minimum distance to all other memebers of the cluster). We call these selected filters "consensus filters".
• best_net_weights.pdf :
  Filter weight vectors of the network that achieved the highest validation accuracy.
• filter_response_differences.pdf :
  Difference in cell filter response between classes for each consensus filter. To compute this difference for a filter, we first choose a filter-specific class, that's the class with highest output weight connection to the filter. Then we compute the average cell filter response (value after the pooling layer) for validation samples belonging to the filter-specific class (v1) and the average cell filter response for validation samples not belonging to the filter-specific class (v0). The difference is computed as v1 - v0. For regression problems, we cannot compute a difference between classes. Instead we compute Kendall's rank correlation coefficient between the predictions of each individual filter (value after the pooling layer) and the true response values. This plot helps decide on a cutoff (filter_diff_thres parameter) for selecting discriminative filters.

training_plots

These plots are generated on the basis of samples used for model training.

• tsne_all_cells.png :
  Marker distribution overlaid on t-SNE map.

In addition, the following plots are produced for each selected filter (e.g. filter i):

• cdf_filter_i.pdf :
  Cumulative distribution function of cell filter response for filter i. This plot helps decide on a cutoff (filter_response_thres parameter) for selecting the responding cell population.
• selected_population_distribution_filter_i.pdf :
  Histograms of univariate marker expression profiles for the cell population selected by filter i vs all cells.
• selected_population_frequencies_filter_i.pdf :
  Boxplot of selected cell population frequencies in samples of the different classes, if running a classification problem. For regression settings, a scatter plot of selected cell population frequencies vs response variable is generated.
• tsne_cell_response_filter_i.png :
  Cell filter response overlaid on t-SNE map.
• tsne_selected_cells_filter_i.png :
  Marker distribution of selected cell population overlaid on t-SNE map.

validation_plots

Same as the training_plots, but generated on the basis of samples used for model validation.

After performing model training once, you can refine the plots with different cutoff values for the selected filters and cell populations. Training does not have to be repeated for refining the plots. The pre-computed results can be used with the option --load_results.

Another relevant argument is --export_selected_cells, which produces a CSV result file for each input FCS file and stores it in outdir/selected_cells. Rows in the CSV result file correspond to cells in the order found in the FCS input file. The CSV result file contains two columns per selected filter, the first indicating the cell filter response as a continuous value and the second containing a binary value resulting from thresholding the continuous cell filter response. This later column is an indicator of whether a cell belongs to the cell population selected by a particular filter.

python ../run_analysis.py -f NK_fcs_samples_with_labels.csv -m NK_markers.csv -i NK_cell_dataset/gated_NK/ -o outdir_NK --group_a CMV- --group_b CMV+ --filter_response_thres 0.3 --load_results --export_selected_cells

Documentation

For additional information, CellCnn's documentation is hosted on http://eiriniar.github.io/CellCnn/