Identifying Differentially Expressed Genes of Zero Inflated Single Cell RNA Sequencing Data Using Mixed Model Score...

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Single-cell RNA sequencing (scRNA-seq) provides high-resolution gene expression data.
• Existing methods often overlook batch effects and zero inflation, leading to biased results.
• Accurate identification of differentially expressed (DE) genes is crucial for biological insights.

Purpose of the Study:

• To propose a novel method, single cell mixed model score tests (scMMSTs), for robust DE gene identification in scRNA-seq data.
• To address the challenges of batch effects and zero inflation inherent in scRNA-seq datasets.
• To evaluate the performance of scMMSTs against existing methods.

Main Methods:

• scMMSTs utilize generalized linear mixed models (GLMMs) to account for batch effects as random effects.
• A weighting strategy is employed to handle zero inflation by calculating observational weights for count data.
• Weighted GLMMs are used for analysis, with score statistics based on mixed Chi-square distributions.

Main Results:

• scMMSTs demonstrated superior performance in identifying DE genes in simulated and real scRNA-seq datasets.
• The method effectively mitigates bias caused by batch effects and zero inflation.
• Comparisons with edgeR-zinbwave and DESeq2-zinbwave showed advantages of scMMSTs.

Conclusions:

• scMMSTs offer an advantageous approach for analyzing zero-inflated scRNA-seq data with batch effects.
• This method serves as a valuable supplement to existing standard DE analysis tools.
• scMMSTs enhance the reliability and accuracy of DE gene discovery in single-cell studies.