VAExprs

VAExprs generates artificial gene expression samples using variational autoencoders to augment limited real observations in biomedical research.

Key Features:

• Variational Autoencoder (VAE): Employs a deep generative model to capture the full probability distribution from observed gene expression data and generate new samples.
• Single-cell RNA-seq Focus: Optimized for generating samples from single-cell RNA-seq data.
• Conditional VAE (CVAE): Enables conditioning on cell types or subpopulations to produce targeted artificial samples rather than completely random ones.

Scientific Applications:

• Data Augmentation: Increases sample size for datasets with few observations due to limited biosamples, high costs, or ethical constraints.
• Robustness Improvement: Enhances reproducibility and robustness of downstream analyses by supplementing real observations with artificially generated samples.
• Targeted Cell Type Generation: Produces specific cell types or subpopulations for focused experimental designs.

Methodology:

VAExprs implements variational autoencoder and conditional variational autoencoder deep generative models to learn the probability distribution of gene expression data and generate new samples.