BSVAE Documentation

BSVAE is a CLI-first workflow for feature-level module discovery in omics data. It combines a Gaussian-mixture variational autoencoder with tools for model selection, network extraction, module assignment, latent analysis, and synthetic benchmarking.

Main Commands

Command

Purpose

bsvae-train

Train a GMMModuleVAE on a features x samples matrix

bsvae-sweep-k

Select the number of modules with held-out validation and optional stability replicates

bsvae-networks

Extract networks, modules, latents, and latent-analysis outputs from a trained model

bsvae-simulate

Generate synthetic datasets, build scenario grids, and benchmark recovery

Data Orientation

Most commands expect an expression matrix in features x samples orientation:

  • rows are feature IDs

  • columns are sample IDs

  • CSV and TSV files use the first column as the feature index

Supported inputs:

  • .csv / .csv.gz

  • .tsv / .tsv.gz

  • .h5 / .hdf5

  • .h5ad with optional anndata

Start Here