This paper addresses the limitations of current GPLVMs for analyzing single-cell RNA sequencing data by introducing an Amortized Bayesian GPLVM (BGPLVM). The model incorporates domain-specific strategies from popular methods such as scVI to effectively handle count data, batch effects, and library size normalization. Experimental results reveal that the proposed model outperforms standard approaches and matches the performance of established methods on real-world datasets, highlighting the significance of specialized modeling in the analysis of single-cell data.
Our proposed model tackles three main aspects of single-cell data: accounting for count data, incorporating batch effect modeling, and normalizing library size through preprocessing.
The model achieves performance comparable to scVI on both simulated and real-world datasets, demonstrating the importance of aligning modeling choices with domain-specific knowledge.
#single-cell-rna-sequencing #bayesian-modeling #dimensionality-reduction #genomics #machine-learning
Collection
[
|
...
]