Document how direct infusion data is handled inFeatureFindingMetabo

[bpben]

Hi, I'm using the tutorial example for [untargeted metabolomics[(https://pyopenms.readthedocs.io/en/latest/feature_detection.html?highlight=FeatureFindingMetabo#metabolomics-untargeted). When I use quant_method "median" in the MassTraceDetection step I get an error when running FeatureFindingMetabo.run: RuntimeError: this method has not been implemented yet. Feel free to complain about it!

My code is below. If I change quant_method to "area", it works fine. I'm assuming that quant_method "median" hasn't been implemented in the FeatureFindingMetabo algorithms, but I wanted to check. If it hasn't been implemented, is it possible to point me to some way I could make a workaround? I'm not familiar with C++, so not sure if I could implement it.

Just as a note - median is suggested as a method for direct injection, which is the source data I'm using at the moment.

# mass trace
mass_traces = []
mtd = MassTraceDetection()
mtd_params = mtd.getDefaults()
mtd_params.setValue("quant_method", 'median') 
mtd.setParameters(mtd_params)
mtd.run(exp, mass_traces, 0)

# elution peak
mass_traces_split = []
mass_traces_final = []
epd = ElutionPeakDetection()
epd_params = epd.getDefaults()
epd.setParameters(epd_params)
epd.detectPeaks(mass_traces, mass_traces_split)
mass_traces_final = mass_traces_split

# feature map
fm = FeatureMap()
feat_chrom = []
ffm = FeatureFindingMetabo()
ffm_params = ffm.getDefaults()
ffm_params.setValue("isotope_filtering_model", "none")
ffm_params.setValue("remove_single_traces", "false") # set false to keep features with only one mass trace
ffm_params.setValue("mz_scoring_by_elements", "false")
ffm_params.setValue("report_convex_hulls", "true")
ffm.setParameters(ffm_params)
ffm.run(mass_traces_final, fm, feat_chrom)

[timosachsenberg]

I think the behavior is probably the same on C++ and python side. e.g., median gets registered here: https://github.com/OpenMS/OpenMS/blob/develop/src/openms/source/FILTERING/DATAREDUCTION/MassTraceDetection.cpp#L53-L54 So maybe it is a combination of parameters that is not supported. @cbielow you have been using FFM quite frequently in the past for direct infusion. Do you see some obvious problem here?

[jpfeuffer]

the problem is here: https://github.com/OpenMS/OpenMS/blob/bf46a11720e76a406addbaafdddaff9f6a48515f/src/openms/source/KERNEL/MassTrace.cpp#L355

[timosachsenberg]

so problem is the smoothing?
can you set the smoothing parameter in FFM to false and try if it works? https://github.com/OpenMS/OpenMS/blob/bf46a11720e76a406addbaafdddaff9f6a48515f/src/openms/source/FILTERING/DATAREDUCTION/FeatureFindingMetabo.cpp#L289 to false

[jpfeuffer]

I don't think it is the smoothing. I think in general nothing is implemented to get the median quantity of a mass trace. It is the getIntensity function

[timosachsenberg]

it is: https://github.com/OpenMS/OpenMS/blob/bf46a11720e76a406addbaafdddaff9f6a48515f/src/openms/source/KERNEL/MassTrace.cpp#L370 but only if not smoothed

[jpfeuffer]

Ah thanks. Correct.

[bpben]

Thanks for the quick responses, appreciate this! If I change the use_smoothed_intensities parameter to "false", it works! That seems to solve it, though haven't looked into the smoothing approach - is it the case that taking the median means the smoothing isn't required?

[timosachsenberg]

I think the error message means that smoothing doesn't make sense for direct infusion data. I will repurpose this issue to a documentation task.

[tapaswenipathak]

What is the documentation todo here?

[timosachsenberg]

might need domain knowledge. @axelwalter may I assign you here?

[axelwalter]

Fixed here. The median quantification method will now be used with smoothing enabled and disabled. It was already documented that the median method should be used for direct infusion data.