NewWave

NewWave models dimensionality reduction and batch effect removal for scRNA-seq data using a negative binomial distribution with flexible dispersion estimation.

Key Features:

• Massive parallelization: Uses shared objects that prevent memory duplication to enable parallel computation.
• Mini-batch approaches: Supports different mini-batch strategies to reduce time consumption.
• Flexible dispersion modeling: Assumes a negative binomial distribution with a dispersion parameter that can be either common across all genes or gene-specific.

Scientific Applications:

• Dimensionality reduction: Reduces the dimensionality of scRNA-seq data for downstream analysis and visualization.
• Batch effect removal: Corrects for batch effects in scRNA-seq datasets while preserving biological variation.

Methodology:

NewWave assumes a negative binomial distribution for scRNA-seq count data with a dispersion parameter that can be either commonwise across genes or genewise.