Diff Iso Search Initial Feedback Action Items
Closed this issue · 5 comments
-
New Isotopic Option: [M+0], but enforce that the max RT is close to the monoisotopic RT (strict RT matching that respects the # scans parameter, or some other RT distance parameter)
Note that it might be possible to accomplish this effect by exposingpeakRtBoundsMaxIntensityFractionin the Peaks dialog, with the goal to explicitly pick peaks that don't descend into the noise, e.g. at the FWHM
It would probably be worth getting a dataset with some specific examples to help resolve this problem - the real issue is when pulling out isotopic peaks, the RT bounds are too wide, the program will sometimes pull out the wrong peak.
A scan-specific approach might be nice, but for now, try the FWHM-based approach, along with options to vary the RT bounds of the peak -
checking the 'extract isotopes' check box and the diff iso check box causes some strange behavior, sometimes the unknown peaks are not being discovered. This might also be related to the number of samples included in each sample set
-
By default, collapse diff isotope search results in the GUI.
-
Option to require the [M+0] isotope is observed in the unlabeled sample set.
-
Peak Detection Dialog: Remove the old 'DotProduct' option, and remember the user setting (don't reset the list every time the Peaks dialog is reopened).
-
Ability to only apply the reproducibility threshold to the labeled sample set (or specify different reproducibility threshold for each set?)
-
For the
[M+0]isotope option, enumerate the quant types as more sophisticated measures than just using the peak area for everything This will have the enormous benefit of using a real peak height, which is important for determining if peaks should be merged between samples. -
Try implementing Calinski-Harabasz index (and think about what to do with F-statistic approach)Pearson correlation coefficient seems to be working OK most of the time, no need to implement this for now. Leave the F-statistic around for the time being, but this should probably be removed at some point.
The peak boundary computation should be revisited - also, the GUI doesn't update in real time, have to reselect the peak group.
X0248 M022A C27H44 [M+H]+ peak group boundaries do not look very good
Even changing to slope threshold of 1 or 10 doesn't appear to change anything
Ultimate goal is quantify C72H44 [M+H]+ in the cerbellum, and figure out why the [M+1] and [M+2] peaks are not always merging together. Potentially propose an option that would force them to go together, even if they are not.
Some of this work has been completed, but much still remains. Moving this off to the next milestone.