Single-Cell Transcriptome Profiling Simulation Reveals the Impact of Sequencing Parameters and Algorithms on...

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Single-cell RNA sequencing (scRNA-seq) analysis lacks methods to quantify cell clustering accuracy due to undefined true cell clusters.
• Transcriptomic heterogeneity is a key feature for defining ground truth in scRNA-seq data.

Purpose of the Study:

• To develop a simulation program (SSCRNA) for scRNA-seq raw data generation.
• To establish a method for quantifying the performance of scRNA-seq clustering algorithms.
• To evaluate the impact of sequencing depth and analytical algorithms on cluster accuracy.

Main Methods:

• Defined a "true" mRNA number matrix based on transcriptomic heterogeneity as ground truth.
• Developed the SSCRNA simulation program mimicking the scRNA-seq data generation process.
• Evaluated consistency between simulated and real scRNA-seq data.
• Quantified the impact of sequencing depth and various normalization/clustering algorithms on cluster accuracy.

Main Results:

• The SSCRNA simulation demonstrated high consistency with actual scRNA-seq data.
• Gaussian normalization was recommended for scRNA-seq data preprocessing.
• K-means clustering showed greater stability compared to K-means combined with Louvain clustering.
• The study quantified the influence of sequencing depth and algorithms on cell cluster accuracy.

Conclusions:

• The developed SSCRNA algorithm accurately simulates the scRNA-seq data generation process.
• This simulation provides a robust framework for comparing normalization and clustering algorithms.
• The findings offer novel insights into optimizing scRNA-seq data analysis and classification.