Using MNN corrected data for per cell pathway scoring

[jiajinlongkang]

Hi MNN group,

I see in the MNN help pages that it is not recommended to use MNN-corrected gene expression matrix for quantitative analysis such as differential gene expression. But I wonder if it is reasonable to use it to calculate pathway scores in each cell (i.e. in each cell, transform the gene expression into pathway scores based on certain pre-defined gene-pathway mappings). This process does not involve any between-cell comparisons. Does this sound reasonable to you?

Thanks you,
Jack

[LTLA]

I would say that the suggestion in the ?fastMNN would still apply to pathway scores; you are still deriving a statistic based on the per-gene corrected profiles, which may not have an obvious interpretation as described here.

Now, it depends on what you want to do with the pathway scores. If all you want is to visualize it on a plot, I'd say go ahead; the corrected values are useful here (and is in fact why they are reported in the first place) to allow people to avoid big jumps in color between batches. However, if you're planning to do actual statistical analyses on them, I would suggest computing the scores from the uncorrected data and applying a blocking factor in your model or test to deal with the batch effect.