P1 / P2 Intensity ~ NAP

[rmflight]

For QC/QA and motivation, we are trying to use the relationship that log(P1 / P2) Intensity - NAP ~ 0.

A couple things we've learned in this:

• Split the grouped_EMF up by labeling patterns, or the differences across all peak comparisons will not ~ 0.
• Use only the best assignments (e-value < 0.1). Higher e-values mean we are more likely to have peaks that don't fit the pattern (sort of the very definition of the higher e-values), and if we don't take all, the labeling might be off or missing.
• Even the fully characterized can be off, so we need to go to scan level. Scan level isn't too bad because we can use matrices to do all scans simultaneously, and therefore only have like 8 - 20 comparisons to do, which is actually really quick.

Generating an across scans metric

• @hunter-moseley suggested that we should use the lowest 10th percentile of values as a summary to compare against full scan-level peaks and xcalibur.
• The best way I can think to do this is first rank the scans by their median or max differences of NAP - Height ratios, and then take the scans in the lowest 10th percentile by that metric, or the lowest 10%.
• To find that lowest 10%, we have three options:
  1. Use only those scans that have all peaks. Alanine for example the labeled 15N only has 63 scans where all the peaks appear.
  2. Use only those peak ratios that appear across all scans. This would mean dropping a peak in the Alanine 15N case.
  3. Remove NA from the median / max calculation and go with it.

[rmflight]

@hunter-moseley: Here is an example that I think demonstrates everything we've suspected about XCalibur and SMIRFE differences, from threonine.

I've gone through and done the log10(NAP1 / NAP2) - log10(Int1 / Int2) for our peaks, and xcalibur peaks (and on a scan-level basis too).

X here is the P1 / P2 index of comparisons, and Y is the absolute values of NAP - Intensity ratios.

We can see that for the peaks where there were more values missing across scans, Xcalibur is objectively doing worse compared to our peaks.

Now that I know what I'm looking for, I can calculate something that tries to summarize this relationship across all of the peaks ...

[hunter-moseley]

Robert,

Very good graph!!

If you order the peaks on the x-axis by n_scan_missing, you could graph maybe all of the peaks and show the relationship very well.

Warm regards, Hunter

On Mon, Mar 28, 2022 at 9:39 PM Robert M Flight @.***> wrote:

@hunter-moseley https://github.com/hunter-moseley: Here is an example that I think demonstrates everything we've suspected about XCalibur and SMIRFE differences, from threonine.

I've gone through and done the log10(NAP1 / NAP2) - log10(Int1 / Int2) for our peaks, and xcalibur peaks (and on a scan-level basis too).

X here is the P1 / P2 index of comparisons, and Y is the absolute values of NAP - Intensity ratios.

[image: threonine_example_differences] https://user-images.githubusercontent.com/1509626/160514642-827ba3a1-a118-4392-a2e1-17c3f89d805d.png

We can see that for the peaks where there were more values missing across scans, Xcalibur is objectively doing worse compared to our peaks.

Now that I know what I'm looking for, I can calculate something that tries to summarize this relationship across all of the peaks ...

— Reply to this email directly, view it on GitHub https://github.com/MoseleyBioinformaticsLab/manuscript.peakCharacterization/issues/12#issuecomment-1081312203, or unsubscribe https://github.com/notifications/unsubscribe-auth/ADEP7B6WITYIZI5YS2HVQ33VCJNOBANCNFSM5RXEAKHA . You are receiving this because you were mentioned.Message ID: <MoseleyBioinformaticsLab/manuscript. @.***>

-- Hunter Moseley, Ph.D. -- Univ. of Kentucky Associate Professor, Dept. of Molec. & Cell. Biochemistry / Markey Cancer Center / Institute for Biomedical Informatics / UK Superfund Research Center Not just a scientist, but a fencer as well. My foil is sharp, but my mind sharper still.

Email: @. (work) @. (personal) Phone: 859-218-2964 (office) 859-218-2965 (lab) 859-257-7715 (fax) Web: http://bioinformatics.cesb.uky.edu/ Address: CC434 Roach Building, 800 Rose Street, Lexington, KY 40536-0093

[rmflight]

OK, so I tried what I read you say, and that does this.

But what I think we are really interested in, is the actual relationship of missing scans and the difference in the xcalibur NAP - Height P1 - P2 to characterized values, which would be this.

This I think should generalize across all of the amino acids in both samples, although most of it will be a really wide spread at zero for all the peaks where there are no scans missing.

[rmflight]

Tagging @jmmitc06 because the plots finally demonstrate things we suspected because the general NAP ~ Height relationship didn't want to work early on in SMIRFE, and we suspected bad averaging was to blame, but the above 3 plots I think really nail it.

And will hopefully nail it across all of the amino acids.

[hunter-moseley]

Robert,

Yes, the last plot is clearest. Agree with your next step to graph all peaks across all amino acids in this plot.

Warm regards, Hunter

On Tue, Mar 29, 2022 at 9:28 AM Robert M Flight @.***> wrote:

Tagging @jmmitc06 https://github.com/jmmitc06 because the plots finally demonstrate things we suspected because the general NAP ~ Height relationship didn't want to work early on in SMIRFE, and we suspected bad averaging was to blame, but the above 3 plots I think really nail it.

And will hopefully nail it across all of the amino acids.

— Reply to this email directly, view it on GitHub https://github.com/MoseleyBioinformaticsLab/manuscript.peakCharacterization/issues/12#issuecomment-1081872712, or unsubscribe https://github.com/notifications/unsubscribe-auth/ADEP7BYJZRXPJUJM3FMOAATVCMAPPANCNFSM5RXEAKHA . You are receiving this because you were mentioned.Message ID: <MoseleyBioinformaticsLab/manuscript. @.***>

-- Hunter Moseley, Ph.D. -- Univ. of Kentucky Associate Professor, Dept. of Molec. & Cell. Biochemistry / Markey Cancer Center / Institute for Biomedical Informatics / UK Superfund Research Center Not just a scientist, but a fencer as well. My foil is sharp, but my mind sharper still.

Email: @. (work) @. (personal) Phone: 859-218-2964 (office) 859-218-2965 (lab) 859-257-7715 (fax) Web: http://bioinformatics.cesb.uky.edu/ Address: CC434 Roach Building, 800 Rose Street, Lexington, KY 40536-0093

[jmmitc06]

Nice, I like those plots and they make sense to me.

[rmflight]

And there we have it. All amino acids from both ECF samples, all assigned formulas with more than 1 peak (so we can do P1 vs P2), comparing the xcalibur NAP - Height to scan-level (uncorrected height) NAP - Height differences.

A and B are for threonine only, and then C is everything.

[hunter-moseley]

Robert,

I think you have your figure!

Warm regards, Hunter

On Tue, Mar 29, 2022 at 4:18 PM Robert M Flight @.***> wrote:

And there we have it. All amino acids from both ECF samples, all assigned formulas with more than 1 peak (so we can do P1 vs P2), comparing the xcalibur NAP - Height to scan-level (uncorrected height) NAP - Height differences.

A and B are for threonine only, and then C is everything.

[image: aa_nap_height_diffs] https://user-images.githubusercontent.com/1509626/160699541-3baeade6-3d7e-4247-856d-5ec91045e936.png

— Reply to this email directly, view it on GitHub https://github.com/MoseleyBioinformaticsLab/manuscript.peakCharacterization/issues/12#issuecomment-1082334169, or unsubscribe https://github.com/notifications/unsubscribe-auth/ADEP7B7KEGE4CGR6HKMEOD3VCNQQDANCNFSM5RXEAKHA . You are receiving this because you were mentioned.Message ID: <MoseleyBioinformaticsLab/manuscript. @.***>

-- Hunter Moseley, Ph.D. -- Univ. of Kentucky Associate Professor, Dept. of Molec. & Cell. Biochemistry / Markey Cancer Center / Institute for Biomedical Informatics / UK Superfund Research Center Not just a scientist, but a fencer as well. My foil is sharp, but my mind sharper still.

Email: @. (work) @. (personal) Phone: 859-218-2964 (office) 859-218-2965 (lab) 859-257-7715 (fax) Web: http://bioinformatics.cesb.uky.edu/ Address: CC434 Roach Building, 800 Rose Street, Lexington, KY 40536-0093

[rmflight]

I think so too.

I think I can also turn this around and generate the intensity version of this for Threonine comparing NAP to intensity of peaks from xcalibur as the motivating figure for the manuscript.