Related Experiment Video
Updated: Sep 23, 2025
Author Spotlight: Vascular Tissue Dissociation and Exploring Single-Cell Subclusters for Targeted Therapy
Published on: January 19, 2024
findPC: An R package to automatically select the number of principal components in single-cell analysis
Haotian Zhuang1, Huimin Wang1, Zhicheng Ji1
1Department of Biostatistics and Bioinformatics, Duke University School of Medicine, Durham, NC 27710, USA.
Summary:
Principal component analysis is widely used in analyzing single-cell genomic data. Selecting the optimal number of principal components (PCs) is a crucial step for downstream analyses. The elbow method is most commonly used for this task, but it requires one to visually inspect the elbow plot and manually choose the elbow point. To address this limitation, we developed six methods to automatically select the optimal number of PCs based on the elbow method. We evaluated the performance of these methods on real single-cell RNA-seq data from multiple human and mouse tissues and cell types. The perpendicular line method with 30 PCs has the best overall performance, and its results are highly consistent with the numbers of PCs identified manually. We implemented the six methods in an R package, findPC, that objectively selects the number of PCs and can be easily incorporated into any automatic analysis pipeline.
Availability And Implementation:
findPC R package is freely available at https://github.com/haotian-zhuang/findPC.
Supplementary Information:
Supplementary data are available at Bioinformatics online.
