Open levinhein opened 2 years ago
Hi @levinhein,
Sorry for the delayed response. I was on vacation. We defined the term meta-analysis vs a standard integrated analysis as no data integration method was used in this paper. As we only had 3 genes for most of the datasets, we instead harmonized the labels based on expert curation of the given cell type labels of each dataset. Thus, we couldn't assess the integrated label set.
I would define the difference as the following: Integrated analysis: Data integration method used to remove batch effects between datasets. Cells are given labels by clustering + re-annotation to ensure labels are consistent. Then any downstream analysis. Meta analysis: Labels are harmonized, but no integration method is used. The downstream analysis is performed in a manner that accounts for batch effects (here: GLMs with batch covariate)
I think whether you mix different tissues depends on your question. There are also integrated atlas studies of fibroblasts or immune cells from different tissues that I wouldn't call a meta-analysis.
I hope that clarifies things!
Hello. Not really an issue but I'd like to reach out with a question as single-cell meta-analysis term is kind of new. I'm curious about what's the difference between the usual Single-cell integrated analysis (method 1) and Single-cell meta-analysis (method 2)?
Is it just the same standard integration analysis but only differs on sample source? Like: method 1 can be integrated scRNA-seq of several same type of tissue sample datasets of one disease method 2 can be integrated scRNA-seq of several different types of tissue datasets of one disease?
Or there's a specialized statistical analysis/package/technique to use in order to call it single-cell meta-analysis?
Scenario question: Can single-cell meta-analysis be also called for a usual integrated scRNA-seq analysis of several PBMC scRNA-seq data but from different but similar diseases (e.g. heart failure + myocarditis + heart attack)?
Thanks so much in advance for any clarification!!!