search

RGLab / flowcap3

1 stars 0 forks source link

Fix transformation of CD27 marker in B-Cell pipeline #2

Closed ramhiser closed 11 years ago

ramhiser commented 11 years ago

Currently, the CD27 observations in the Stanford data set are not being transformed properly. The range of values for the other centers is 0 to 4.5, whereas Stanford is -100 to 262,143.0.

Here is a kernel density estimate for the first Stanford sample conditional on CD19+CD20+. The source of the poor transformation is likely in the munge code.

Screen Shot 2013-04-11 at 3 59 46 PM

I noticed the issue when the prior was elicited and lead to terrible gating. Notice the range of values in the distribution of peaks (they are all collapsed into a single vector across all samples):

Screen Shot 2013-04-11 at 3 56 13 PM

ramhiser commented 11 years ago

After stepping through the munge scripts that compensate and transform the data for the Bcell panel, it turns out this script is ignoring the CD27 marker for Stanford when compensating. This is because the Excel file provides no filename for this marker (as well as some other markers related to other panels).

This is then likely related to a conversation Greg and I had earlier in the week. Specifically, Greg noted in an email:

Meena clarified that they used generic color controls for the panels, so you can use the color (PE-Cy7) tube for all other markers using that color in other panels, rather than markers specific compensation.

I will update the munge script to use a generic color control if no filename is provided for a marker.

ramhiser commented 11 years ago

After the changes to the munge scripts, the scale of the Stanford data is 0 to 4.5, which is similar to the other centers.