A Reproducibility Focused Meta-Analysis Method for Single-Cell Transcriptomic Case-Control Studies Uncovers Robust...

Area of Science:

• Genomics
• Neuroscience
• Infectious Disease

Background:

• Single-cell RNA sequencing (scRNA-seq) studies generate large datasets for disease research.
• Reproducibility of differentially expressed genes (DEGs) across studies is crucial for reliable biomarker discovery.
• Previous analyses showed variable reproducibility of DEGs in neurological and infectious diseases.

Purpose of the Study:

• To evaluate the reproducibility of DEGs in Alzheimer's (AD), Parkinson's (PD), Schizophrenia (SCZ), and COVID-19 scRNA-seq studies.
• To develop and validate a novel meta-analysis method for enhancing DEG reproducibility and predictive power.
• To identify robust DEGs and associated biological pathways for these diseases.

Main Methods:

• Assessed reproducibility of DEGs from published AD, PD, SCZ, and COVID-19 scRNA-seq datasets.
• Developed SumRank, a non-parametric meta-analysis method based on relative differential expression rank reproducibility.
• Compared SumRank with dataset merging and inverse variance weighted p-value aggregation methods.

Main Results:

• Individual PD and COVID-19 DEG datasets showed moderate predictive power (AUC=0.77, 0.75); AD and SCZ showed poor power (AUC=0.68, 0.55).
• SumRank improved DEG predictive power across diseases (AUC=0.88, 0.91, 0.78, 0.62).
• SumRank-derived DEGs exhibited higher specificity and sensitivity than other methods and revealed known/novel pathways, including validation of BCAT1 in AD mouse oligodendrocytes.

Conclusions:

• The SumRank method significantly enhances the reproducibility and predictive power of DEGs from scRNA-seq studies.
• This approach facilitates more reliable discovery of disease-associated genes and pathways in complex diseases.
• Understanding factors influencing study reproducibility is vital for future experimental design in transcriptomics.