bmVAE is a variational autoencoder method for clustering single-cell mutation data.

• Python 3.8+.

First, download bmVAE from github and change to the directory:

git clone https://github.com/zhyu-lab/bmvae
cd bmvae

Create a new environment named "bmvae":

conda create --name bmvae python=3.8.13

Then activate it:

conda activate bmvae

Use pip to install the requirements:

python -m pip install -r requirements.txt

Now you are ready to run bmVAE!

bmVAE clusters cells into distinct subpopulations and infer genotypes of each subpopulation based on single-cell binary data.

Example:

python bmvae.py --input testdata/example.txt --output testdata

The SNVs of single cells are denoted as a genotype matrix. Each row defines the mutational states of a single cell, and each column represents a mutation. Columns are separated by tabs. The genotype matrix is binary.

The entry at position [i,j] should be

• 0 if mutation j is not observed in cell i,
• 1 if mutation j is observed in cell i, or
• 3 if the genotype information is missing

The output directory is provided by users.

The genotypes of subpopulations are written to a file with name "clusters.txt".

The cell-to-cluster assignments are written to a file with name "labels.txt".

The estimated FPR and FNR are written to a file with name "para.txt".

• --input <filename> Replace <filename> with the file containing the genotype matrix.

• --output <string> Replace <string> with the output directory.

• --Kmax <INT> Set <INT> to a positive integer. This specifies the maximum number of clusters to consider. Default value is set to N/10 (N denotes the number of cells).

• --seed <INT> Set <INT> to a non-negative integer. This specifies the seed for generating random numbers. Default value is 0.

• --epochs <INT> Set <INT> to a positive integer. This specifies the number of epoches to train the VAE. Default value is 250.

• --batch_size <INT> Set <INT> to a positive integer. This specifies the batch size for training the VAE. Default value is 64.

• --lr <Double> Set <Double> to a positive real number. This specifies the learning rate for training the VAE. Default value is 0.0001.

• --dimension <INT> Set <INT> to a positive integer. This specifies the dimension of latent space of the VAE. Default value is 3.

If you have any questions, please contact [email protected].