Use/example for MS:4000156 (area under TIC RT quantiles) as quality criteria

[jorainer]

Picking up on a small discussion/question we had at the mzQC session in Tuebingen. We (in xcms-based LC-MS data preprocessing) generally create TICs and base peak chromatograms (BPC) to evaluate how similar the samples, respectively their chromatographic signals, are (within the same experiment). We use this to spot potential problematic samples/runs and ensure that e.g. QC samples are similar etc. Also, we use this to evaluate the performance of the retention time alignment step.

While this can be visually evaluated, we obviously also want to have a quantitative assessment. To do that we used to bin the BPCs (in bins of e.g. 2 seconds, or 5 seconds) and then used heatmaps and hierarchical clustering or correlation matrices to get a numeric estimate. After digging through the various mzQC measures I stumbled over the area under TIC RT quantiles measure - and we now use this, i.e. instead of binning the data in x second bins we bin the signal by e.g. 10 or 20 quantiles.

As an example: below are BPC of a small experiment, with one of the samples showing a retention time shift.

A heatmap of area under BPC RT quantiles for that data is shown below. The samples/runs are along the y-axis and the retention time bins (19 quantiles) along the x-axis. The data is not bad, but the sample replicates do not group as expected.

The same plots after retention time alignment:

the shift was properly adjusted, and also the runs cluster as expected based on the area under BPC RT quantiles:

We include this quality check now regularly in our analyses and will also add an example to our Metabonaut resource. My questions now would be:

1. would it be possible to create a new "area under BPC RT quantiles" mzQC term?

2. would this be something interesting to report in your paper (or supplement)? As an example/use case for this mzQC term? If yes, we would need to look for other (publicly available) input data to provide a completely reproducible example.

Happy to discuss :)

[bittremieux]

Thanks for the detailed information.

1. Yes, we can certainly have this complementary CV term. It seems that you're already quite familiar with the CV, so do you mind drafting an initial description to be included in there?
2. As a single QC metric, we probably don't need to explicitly cover it in the manuscript. However, the more examples we have the better, e.g. on the website. If you have some documentation on Metabonaut, we should certainly link to it.

[jorainer]

thanks for the reply - sounds good. I'll request a new CV term (soonish) and will provide the link once we have an example in Metabonaut.

[bittremieux]

@jorainer Does this work as a CV term for your metric?

[Term]
id: MS:4000XXX
name: base peak chromatogram area quantiles
def: "Summarizes the chromatographic signal distribution by integrating the base peak chromatogram (BPC) across equally sized retention time quantiles within a single LC–MS run. Consistency in these profiles across samples indicates similar chromatographic behavior and proper retention time alignment."
is_a: MS:4000XXX ! chromatographic performance metric
relationship: part_of_workflow_stage MS:4000XXX ! chromatography stage
relationship: depends_on_data_type MS:4000XXX ! raw acquisition data
relationship: has_measurement_scope MS:4000XXX ! run level
relationship: applies_to_acquisition_mode MS:4000XXX ! acquisition mode independent
relationship: has_quality_directionality MS:4000XXX ! context dependent
relationship: has_value_type MS:4000XXX ! tuple
relationship: has_value_concept STATO:0000291 ! quantile
relationship: has_units MS:1000043 ! intensity unit
relationship: has_value_type xsd:float

[jorainer]

yes @bittremieux ! that would be great!