I am currently working with single-cell RNA sequencing (scRNA-seq) data from a specific cell type under six different dietary conditions in two age groups (12 integrated datasets). My aim is to compare these conditions to understand which diets induce similar gene expression patterns and how they might relate to aging markers. However, I've encountered challenges in my analysis.
Batch Effect Confusion: Initially, I did not observe any obvious batch effects in PCA, UMAP, and TSNE. However, when I looked at the expression of aging marker genes, I noticed that diets sequenced in the same batch appeared more similar. This suggests there might be hidden batch effects affecting my data.
Pseudo-bulk Analysis Attempt: To address this, I tried a pseudo-bulk analysis where I aggregated gene expression from cells across all diets and applied batch correction using the ComBat algorithm. Surprisingly, this seemed to eliminate the batch effects in hierarchical clustering. However, I then realized that differences in the number of cells per diet were influencing the aggregated gene expression, leading to biased results.
Dilemma with Aggregation: At this point, I'm unsure about the best approach. Should I continue with pseudo-bulk analysis, perhaps using average expression instead of simple aggregation? Or are there alternative methods to address batch effects and accurately compare diets while accounting for variations in cell numbers?
Any guidance or recommendations would be greatly appreciated!
Has data been generated in the same batch, so same day, in the same 10x runs?
How similar, exactly? Can you post any of your expression data?