event ratio correction and cell number correction

[Harshangda Karan Puri]

Hi,

This might be a very naive question, but kindly help me as I am still learning. I am comparing isoform expressing and usage in three populations AGM, YSb, YSm. Each of these three populations consist of cells from two clusters from each trajectory of a previously done trajectory analysis i.e. AGM population has cells from AGM1 and AGM2 subcluster which together I am considering as single population AGM. Similarly, YSb has YSb1 and YSb2 subclusters which I am together considering as single population YSb. And YSm has YSm1 and Ysm2 subclusters that I am considering as single population YSm.

The problem is I have 40 AGM1 cells and 20 AGM2 cells, total 60 AGM cells.

98 YSb1 cells and 100 YSb2 cells, total 198 YSb cells.

96 YSm1 and 92 YSm2, total 188 YSm cells.

My question is do I need to correct for event ratio between the AGM1 and AGM2 clusters so that they are approx 1:1 just like YSb1:YSb2 and YSm1:YSm2? If so, how can this be done using IsoformSwitchAnalyzeR?

Also, is it a problem if I have 60 AGM, 198YSb, and 188YSm cells?  Do I need corrections for this? if yes, how can this be done?

Thanks,

Harsha

[Kristoffer Vitting-Seerup]

IsoformSwitchAnalyzeR will use generalized linear modeling to handle this, meaning it will just consider the mean expression in each category. The only problem this could cause is that you naturally have more statistical power in the condition with more cells and hence you could find smaller differences. But in practice I don't think that will be a problem since IsoformSwitchAnalyzeR also employs an effect-size cutoff per default 🙂