Deconomix Documentation

The deconomix package provides functions for (Adaptive) Digital Tissue Deconvolution. It is split into two collections of functions, deconomix.utils and deconomix.methods, where the first provides useful utilities around the usual deconvolution workflow and the latter provides the actual models and training routines.

Modules

In the article, we divided the functionalities of Deconomix semantically into three different modules, which are realizable with the Python functions we provide in the package. However, the functions have a broad spectrum of use-cases overall, which we want to keep for advanced users. Therefore we want to give a summary on how to use the modules described in the article here:

Module 1: Learning Gene Weights:

X_ref, Y_train, C_train = deconomix.utils.simulate_data(scRNA_df, n_mixtures=10000)
module1 = deconomix.methods.DTD(X_ref, Y_train, C_train)
module1.run()
gene_weights = module1.gamma

Module 2: Cellular Composition (with hidden background)

module2 = deconomix.methods.ADTD(X_ref, test_bulks_df, gene_weights, C_static = False, Delta_static = True)
module2.run()
Cellular_Contributions   = module2.C_est
Hidden_Contributions     = module2.c_est
Hidden_Consensus_Profile = module2.x_est

Module 3: Gene Regulation

# Search hyperparameter with deconomix.methods.HPS() or determine otherwise, e.g. 1e-6
module3 = deconomix.methods.ADTD(X_ref, test_bulks_df, gene_weights, C_static=True, Delta_static = False, lambda2 = 1e-6)
# This also updates the cellular contributions slightly.
Cellular_Contributions = module3.C_est
Hidden_Contributions   = module3.c_est
Hidden_Consensus_Profile = module3.x_est
Gene_Regulation_Factors = module3.Delta_est

Installation

You can install the package using the official PyPI repositories:

pip install deconomix

Alternatively, clone the repository and install from the directory:

pip install git+https://github.com/Deconomix/Deconomix.git