search

MoseleyBioinformaticsLab / manuscript.peakCharacterization

0 stars 1 forks source link

failures for some samples using new frequency model #49

Closed rmflight closed 2 years ago

rmflight commented 2 years ago

There are still 9 lung samples (better than 30 previously) that don't seem to want to complete when switching to the new frequency model.

I would like to run them in SCPC proper and see if they complete using either the old or new frequency model, so we can posit whether it is the old implementation or the new frequency model.

rmflight commented 2 years ago

Thus far, they are all completing without error under the new SCPC package with the default (4 component) frequency model. And then also completing under the old FTMS peak characterization with the 3 component model.

So my hypothesis is that there is some kind of weird interplay between the old FTMSPC code and the 4 component frequency model. Probably due to some assumptions baked into the scan filtering that get broken.

rmflight commented 2 years ago

All the previously not run samples ran in both the new SCPC :package: with 4 component model, and the old FTMS :package: with a 3 component model.