pc.scalefactors are not identical to results from positiveQC

[chunxuan-hs]

I am a bit confused about the postive control steps.

In the codes

dta <- processNanostringData(nsFiles       = path_data,
                             sampleTab     = path_meta,
                             normalization = "nSolver",
                             idCol         = "RCC_Name",
                             output.format = "list")

It gives warning that

Identified positive scale factor outside recommended range
                    (0.3-3).
Check samples prior to conducting analysis.

And indeed in the dta$pc.scalefactors, there are small values.

However, when I run the positiveQC, I got the different scale values.

posQC <- positiveQC(dta)

Actually I expect them to be identical. Any insights are appreciated. Thanks!

Update: The R codes show that the scale factor in positiveQC are calculated on log scale, while scale factor in processNanostringData is calculated in raw scale? Regarding the recommended range (0.3-3), which approach is better?

[calebclass]

Thank you for raising this issue! NanoString recommends using the geometric mean of the raw counts to compute the scale factor. That means processNanostringData with normalize_pos_controls is working properly, while positiveQC is incorrect.

I will work on pushing a fix to positiveQC soon, but please use the scale factors from processNanostringData for now (ideally, I should get positiveQC to use normalize_pos_controls, but I'm not sure how easy that will be). I believe positiveQC is calculating R-squared correctly for Observed vs. Expected, because NanoString recommends for this statistic to be calculated from log2 values. But I will double-check to make sure.

[chunxuan-hs]

Many thanks for the updates!