API Reference

Below is the class and function reference for scDeepSort. Notice that the package is under active development, and some features may not be stable yet.

DeepSortPredictor

# Class Definition
DeepSortPredictor(species,
                  tissue,
                  file_type='csv',
                  unsure_rate=2.)

  • species: The species of cells, human or mouse.

  • tissue: The tissue of cells. For the detailed list of supported tissue of our model, please refer to GitHub Wiki page.

  • file_type: The format of data file, csv or gz. csv for .csv files and gz for .gz files. For details, please refer to R script.

  • unsure_rate: The multiplier for unsure threshold (computed as unsure_rate / num_classes) to cast the type of cell to the unsure type, default to 2.0. Set it as 0 to exclude the unsure type.

# Class Method
DeepSortPredictor.predict(input_file,
                          save_path=None) -> pandas.DataFrame

  • input_file: The file path for test dataset.

  • save_path: The destination for saving predictions. Save results to disk if path provided.

Example

from deepsort import DeepSortPredictor
# define the model
model = DeepSortPredictor(species='human',
                          tissue='Brain')
# use the trained model to predict
test_files = ['/path/to/human_brain_test_data_1.csv', '/path/to/human_brain_test_data_2.csv']
for test_file in test_files:
    model.predict(test_file, save_path='results', model_path='model_save_path')

DeepSortClassifier

# Class Definition
DeepSortClassifier(species,
                   tissue,
                   dense_dim=400,
                   hidden_dim=200,
                   batch_size=256,
                   dropout=0.1,
                   gpu_id=-1,
                   file_type='csv',
                   learning_rate=0.001,
                   weight_decay=5e-4,
                   n_epochs=300,
                   n_layers=1,
                   threshold=0,
                   num_neighbors=None,
                   exclude_rate=0.005,
                   random_seed=None,
                   validation_fraction=0.1)

  • species: The species of cells, human or mouse.

  • tissue: The tissue of cells. For the detailed list of supported tissue of our model, please refer to GitHub Wiki page.

  • dense_dim: The initial dimension of node embedding for cells and genes. Default to 400.

  • hidden_dim: The hidden dimension of Weighted Graph Aggregator Layer. Default to 200.

  • batch_size: The number of samples per batch. Default to 256.

  • dropout: The dropout rate for the output representation of Weighted Graph Aggregator Layer, default to 0.1.

  • gpu_id: The GPU id for training and testing. -1 for CPU. Default to -1.

  • file_type: The format of data file, csv or gz. csv for .csv files and gz for .gz files.

  • learning_rate: The learning rate of optimizer. Default to 0.001.

  • weight_decay: The weight decay of optimizer. Default to 0.00004.

  • n_epochs: Maximum number of epochs. Default to 300.

  • n_layers: The number of layers. Default to 1.

  • threshold: The weight threshold for edges between cell nodes and gene nodes. Default to 0.

  • num_neighbors: The number of sampled neighbors per nodes in training time. If None, all the neighbors will be sampled.

  • exclude_rate: Exclude a class if the portion of this class is less than exclude_rate. Default to 0.005.

  • random_seed: For reproducibility. Fixed if given. Default to None.

  • validation_fraction: The proportion of training data to set aside as validation set for early stopping. Must be between 0 and 1. Default to 0.1.

# Class Method
DeepSortClassifier.fit(files,
                       save_path=None)

  • files: The file path for training datasets. We assume files in the form of list of (data_file, celltype_file).

  • save_path: The destination for saving models.

# Class Method
DeepSortClassifier.predict(input_file,
                           model_path,
                           save_path=None,
                           unsure_rate=2.,
                           file_type='csv')  -> pandas.DataFrame

  • input_file: The file path for test dataset.

  • model_path: The path for loading saved models.

  • save_path: The destination for saving predictions. Save results to disk if path provided.

Example

from deepsort import DeepSortClassifier
# define the model
model = DeepSortClassifier(species='human',
                           tissue='Brain',
                           dense_dim=50,
                           hidden_dim=20,
                           gpu_id=0,
                           n_layers=2,
                           random_seed=1,
                           n_epochs=20)
train_files = [('/path/to/human_brain_data_1.csv', '/path/to/human_brain_celltype_1.csv'),
               ('/path/to/human_brain_data_2.csv', '/path/to/human_brain_celltype_2.csv')]
test_files = ['/path/to/human_brain_test_data_1.csv', '/path/to/human_brain_test_data_2.csv']
# fit the model
model.fit(train_files, save_path='model_save_path')
# use the saved model to predict
for test_file in test_files:
    model.predict(test_file, save_path='results', model_path='model_save_path')