I am interested in comparing gene expression between adult and aged mice in specific brain regions using the following two datasets:
10x-scRNAseq-aged-adult
MERFISH-C57BL6J-638850 with Imputed Genes + Reconstructed Coordinates
Since the 10x scRNA-seq dataset does not contain spatial information, I am considering using the MERFISH dataset to identify region-specific marker genes. I would then use these markers to infer the anatomical origins of cells in the 10x scRNA-seq dataset and investigate age-related changes in gene expression within specific brain regions.
I have a few questions:
Are the CPM values directly comparable between these two datasets, particularly for genes imputed in the MERFISH dataset?
Are there differences in normalization or expression scaling that should be considered when comparing these datasets?
If direct comparison of CPM values is not appropriate, would it be reasonable to use MERFISH for identifying region-specific markers and then perform adult-versus-aged comparisons entirely within the 10x scRNA-seq dataset?
I would greatly appreciate any guidance on this approach.
Regarding to your question. Yes, the imputed MERFISH gene expression is based on logCPM of reference RNAseq data. So they should be comparable, but I wouldn’t recommend comparing them with aged 10X scRNAseq data directly, because the distribution has changed substantially. The imputed value are very smooth, but the original 10X scRNAseq data have many drop outs, so depends on the statistical test you performed, they may caused some issues. If the idea to to visualize which brain region has the most gene expression differences, I would recommend integrate MERFISH data with the adult and age data directly, and impute adult and age gene expression for each MERFISH cells, and compute the difference of the imputed values that way. Special care needs to be considered to QC MERFISH cells properly, and maker sure integration is tight, and cell type coverage are roughly comparable between the age group.
Thank you very much for your detailed and helpful response.
Your explanation of the differences in expression distributions between imputed MERFISH and original 10x scRNA-seq data was very informative.
I am particularly interested in your suggestion to integrate the adult and aged scRNA-seq datasets with the MERFISH reference separately, impute gene expression for each MERFISH cell, and then compare the imputed values.
I will explore this approach while paying careful attention to MERFISH cell quality control, integration accuracy, and comparable cell-type coverage between age groups.
One thing I’d be careful about is using the same marker genes both to assign cells to a brain region and then to test for age-related expression changes.
I wonder if that could introduce some circularity, especially when the regional labels are inferred rather than directly observed.
Maybe it would be worth checking whether the main results hold when the region assignments are based on a separate marker set, or when those genes are excluded from the downstream comparison.
It wouldn’t solve the differences between the datasets, but it might make the age-related findings a bit easier to interpret.