Manual Integration severely under-estimating quant on some QQQ peaks

[cmsand]

Maven version: 2.1.0 Dataset: 20230403_CD38-inhibition-knockdown-NAADP_Rerun-2 Compound: NR1 [M+H]+ Samples:

• CD38-inhibition-knockdown-NAADP-Rerun-2-Q0004_M027A_09_THP1M0-DMSO.mzML
• CD38-inhibition-knockdown-NAADP-Rerun-2-Q0004_M027A_18_THP1M0-DMSO.mzML
• CD38-inhibition-knockdown-NAADP-Rerun-2-Q0004_M027A_29_THP1M0-DMSO.mzML

Manual integration using the "shift-drag" method was done on this peak group because of one RT outlier sample (not shown), but doing so significantly mis-quantified the injection 09 sample above. No other samples were seriously affected by a first-pass look. Injections 18 and 29 shown for comparison. Screenshots showing the problem below.

Peak integration from qqq results table (sample order is 09, 18 ,29. Sample 09 is dark orange):

After manual integration (sample order is 09, 18 ,29. Sample 09 is dark orange):

[cmsand]

It's worth noting that the manually-integrated quant on bookmarks passed through to the analysis stage even with smoothedPeakAreaCorrected selected as the quant type. This is the behavior we want usually, just the issue with how manual integration handled this peak in this case.

[cmsand]

Maven file with issue added here https://drive.google.com/drive/folders/1Xr6mp4FdO7PInfuk6l520Ht4APmlvqEa

Dataset here: /massspec-group/qqq_datasets/20230403_CD38-inhibition-knockdown-NAADP_Rerun-2

[cmsand]

Another example where a peak was missed entirely:

QQQ Results Table Integration:

Manual Integration:

[PMSeitzer]

Marking as "not reproducible" for now, hoping to be able to figure out how exactly to trigger the bug condition after more investigation. I couldn't see anything obvious in the code to try and trigger the bug condition.