The article discusses a novel approach to Gaussian Process Latent Variable Models (GPLVMs) tailored for scRNA-seq data analysis. It emphasizes the incorporation of domain knowledge via customized kernel designs, which address batch effects and cell-cycle influences. These adjustments offer improved interpretability of latent spaces while requiring less training data. The proposed model shows significant enhancements over standard Bayesian GPLVM methods, highlighting its practical implications in biological data analysis, especially in understanding latent variables related to cell cycles and batch effects.
"A key benefit of using GPLVMs is that we can encode prior information into the generative model, especially through the kernel design, allowing for more interpretable latent spaces and less training data."
"In order to correct for confounding batch effects through the GP formulation, Lalchand et al. (2022a) proposed the following kernel structure with an additive linear kernel term to capture random effects."
Collection
[
|
...
]