Install DCA via pip or conda
masterYou can install the Deep Count Autoencoder (DCA) using either pip or conda.
Using pip:
pip install dcaUsing conda:
conda install -c bioconda dcaDeep Count Autoencoder (DCA)
repository·master·Indexed 42 days ago
https://github.com/theislab/dcaA tool for denoising scRNA-seq data using a deep autoencoder with a zero-inflated negative binomial (ZINB) loss function to account for count structure, overdispersion, and sparsity.
What's inside DCA
Understand DCA output files
masterDCA generates several TSV files in the specified output directory representing different parameters of the Zero-Inflated Negative Binomial (ZINB) distribution and the latent space:
File Description mean.tsvThe main output representing the mean parameter of the ZINB distribution. Dimensions match the input (excluding zero-expression genes/cells). Formatted as gene x cellmatrix.mean_norm.tsvLibrary size-normalized expressions. Uses the same method as Scanpy's normalize_totalfunction.pi.tsvDropout probabilities for each cell and gene. dispersion.tsvDispersion parameters for each cell and gene. reduced.tsvThe hidden representation (latent space) of each cell. By default, this is a 32-dimensional space representing bottleneck neuron activations. Perform hyperparameter optimization
masterTo perform a hyperparameter search instead of a standard run, include the
--hyperoption in your command line invocation.dca matrix.csv results --hyperRun DCA from the command line
masterTo denoise scRNA-seq data using the DCA command-line interface, use the following syntax:
dca <input_matrix> <output_directory>Requirements:
<input_matrix>: A CSV or TSV-formatted raw count matrix where genes are in rows and cells are in columns.- Mandatory: Both cell and gene labels must be present in the input file.
<output_directory>: The directory where results will be saved.
dca matrix.csv results