Score intra-row purity by MS2 similarity¶
Feature list methods → Processing → Score intra-row purity by MS2 similarity
Compares all MS/MS spectra that belong to the same feature list row against each other and scores their pairwise modified cosine similarity. A row whose spectra disagree is suspicious: the row either aggregates fragmentation spectra of co-eluting isomers or isobars, or - on an aligned feature list - it aligned features of different compounds.
The result is a quality measure per row that is independent of any annotation, so it can be used before or without spectral library search.
Where this is useful¶
- MALDI MS2 pairing and the SIMSEF workflow. In MALDI MS2 scan pairing (SIMSEF) a single feature collects fragmentation spectra acquired on many different MALDI spots. Those spectra should be identical apart from noise. A low intra-row similarity indicates that the imaged spots do not contain the same compound at that m/z, or that the isolation was chimeric on part of the measurement area. Scoring the row purity is therefore a natural QC step after running the SIMSEF scheduler and pairing the acquired MS2 spectra, and before generating a spectral library from the data.
- Aligned feature lists. Low scores point to rows where features of different compounds were aligned.
Results¶
Each processed row receives an Intra-row MS2 similarity column containing the summary statistics
(average, minimum, maximum, and number of compared pairs) of all similarity values of that row,
together with a group label that bins the average score into < 0.4, [0.4, 0.6), [0.6, 0.85),
and >= 0.85. Sorting the feature table by this column brings the most suspicious rows to the top.
Tip
Rows with only a single fragmentation spectrum cannot be scored and stay empty. This is expected and does not indicate a problem.
Parameters¶
Feature lists¶
The feature list to process.
Export to file (Optional)¶
Additionally writes a CSV file with the columns row_id and cosine_similarity, containing one
line per compared spectrum pair. Useful to inspect the score distribution outside of mzmine.
Disabled by default.
m/z tolerance (MS2)¶
Tolerance for matching signals between two MS/MS spectra. Default is 0.003 Da or 10 ppm.
Tip
Keep a reasonable absolute tolerance. A purely relative tolerance such as 5 ppm is too narrow to cover small neutral losses at low m/z values, for example around m/z 18.
Minimum signals¶
Minimum number of signals a spectrum must contain to be compared. Low quality scans below this threshold are skipped. Default is 4.
Signal filters¶
Filters applied to each spectrum before comparison:
- Remove residual precursor m/z - removes signals around the precursor, including its isotopes. Strongly recommended, because precursor signals always match and would inflate the similarity. Enabled by default with a window of m/z 10.
- Crop to top N signals - keeps only the most abundant signals. Default is 250.
- Signal threshold (intensity filter) - above this number of signals, the spectrum is reduced to the top signals that make up the configured intensity percentage. Default is 50.
- Intensity filter at >N signals - the intensity fraction retained by the filter above. Default is 98%.
Split by collision energy¶
If enabled (default), only spectra acquired with exactly the same collision energy are compared. Disable this only if your acquisition uses varying or ramped energies that should still be treated as comparable.
Warning
With this option disabled, spectra of different collision energies are compared directly, which lowers the similarity of perfectly pure rows.