Accelerated dimensionality reduction of single-cell RNA sequencing data with fastglmpca.

Area of Science:

• Computational biology
• Bioinformatics
• Statistical genetics

Background:

• Principal Component Analysis (PCA) is widely used for dimensionality reduction in single-cell RNA sequencing (scRNA-seq) data.
• Traditional PCA methods face challenges when applied to count data, leading to the development of Poisson Generalized Linear Model PCA (GLM-PCA).
• Fitting GLM-PCA can be computationally intensive, posing a barrier for large-scale scRNA-seq analyses.

Purpose of the Study:

• To address the computational challenges associated with fitting GLM-PCA for scRNA-seq data.
• To introduce a novel, efficient algorithm for dimensionality reduction of count data.
• To provide an accessible R package for implementing the new algorithm.

Main Methods:

• Development and application of a new algorithm termed Alternating Poisson Regression (APR).
• Comparison of APR's performance against existing GLM-PCA fitting algorithms.
• Implementation of APR in the fastglmpca R package for user accessibility.

Main Results:

• APR demonstrates superior fit quality compared to existing algorithms for GLM-PCA.
• APR achieves dimensionality reduction in less computational time.
• The algorithm is memory-efficient and amenable to parallel processing, suitable for large scRNA-seq datasets.

Conclusions:

• Alternating Poisson Regression (APR) provides a computationally efficient and effective solution for dimensionality reduction of scRNA-seq count data.
• The fastglmpca R package offers a practical tool for researchers to apply advanced PCA methods to their scRNA-seq datasets.
• APR enhances the scalability and accuracy of analyzing high-throughput sequencing data.