Command-line interface

bulkdgd is structured as an importable Python package.

However, a command-line interface is provided for some of the most common tasks bulkdgd is used for.

This interface consists of a series of executables installed together with the package:

• bulkdgd_get_recount3 allows the seamless retrieval of RNA-seq data and their associated metadata from the Recount3 platform.

• bulkdgd_get_genes allows the creation of custom lists of genes to use with the bulkdgd model.

• bulkdgd_preprocess_samples allows the preprocessing of samples’ data before using them with the bulkdgd model.

• bulkdgd_find_representations allows finding the best representations in the latent space defined by the bulkdgd model for a set of new samples.

• bulkdgd_dea allows performing differential gene expression analysis between a set of samples and their ‘normal’ counterparts found by the bulkdgd model.

• bulkdgd_reduction_pca / bulkdgd_reduction_kpca / bulkdgd_reduction_mds / bulkdgd_reduction_tsne / bulkdgd_reduction_umap allows performing dimensionality reduction analyses and plotting the results.

• bulkdgd_find_probdens allows finding, for a given set of representations, the probability density of each representation for each component of the Gaussian mixture model that defines the bulkdgd model’s latent space.

• bulkdgd_train allows training the model.