Consumer check

Consumer check

A severity is a claim about what a difference costs a reader. This page holds the evidence for those claims: one defect at a time injected into a real file, and what mzdiff, mzValidate, and a search engine (Sage) each made of it.

How it was run

tools/consumer.py (Development) edited the same 200 spectra in each case: rank-1 target matches at 1% FDR in the reference, spread over the run. The reference was the ThermoRawFileParser mzML of a HeLa run on an Orbitrap Eclipse (PRIDE PXD016921, 36,341 spectra, about 11,290 target matches at 1% FDR), searched against the human proteome with Sage 0.15.0-beta.2.

Sage's outcome for the 200 spectra, in the Sage column: fatal (the run fails or reads nothing), lost (spectra vanish from its results), wrong (another peptide or charge, or a lost 1% match), shifted (the same matches with other scores, times, or masses), or none.

Results

Defect in B	mzDiff	mzValidate	Sage
spectrum removed	critical	error	200 lost
spectrum written twice	critical	error	200 reported twice
charge changed	critical	valid	110 lost, 90 wrong
charge removed	critical	valid	6 wrong, 4 shifted
charge written 2.0	critical (form:integer)	error	fatal
precursor m/z 1 isotope off	critical	valid	1 lost, 70 wrong, 129 shifted
precursor m/z 10 ppm off	critical	valid	4 lost, 191 wrong
precursor m/z 0.5 ppm off	minor (ppm)	valid	200 shifted
precursor m/z removed	critical	valid	2 wrong, 43 shifted
isolation target 1 m/z off	critical	valid	none
isolation offsets doubled	major	valid	none
precursor intensity removed	major	valid	none
spectrumRef removed	major	valid	none
collision energy changed	major	valid	none
ms level 2 written as 1	critical	valid	200 lost
ms level written 2.0	critical (form:integer)	error	fatal
centroid flag removed	critical	error	fatal
centroid changed to profile	critical	valid	fatal
total ion current set to 0	critical	valid	200 lost
total ion current doubled	major	valid	none
filter string changed	major	valid	none
injection time doubled	major	valid	none
ms level through a group	major (form:paramGroup)	valid	200 lost
time unit dropped	critical	error	fatal
time in minutes, labelled s	critical	valid	1 wrong, 199 shifted
time converted to seconds	same	valid	none
time one minute later	critical	valid	200 shifted
time removed	critical	valid	2 wrong, 198 shifted
intensity array removed	critical	error	fatal
most intense peak removed	critical	valid	3 wrong, 95 shifted
peaks in falling m/z order	major (array:peaks)	valid	6 wrong, 192 shifted
intensities stored as 32-bit	minor (precision)	valid	none
m/z stored as 32-bit	minor (precision)	valid	none
arrays uncompressed	same	valid	none
m/z array in Numpress	major (form:numpress)	error (wrongly)	200 lost
Base64 broken across lines	same (not visible yet)	valid	fatal
defaultArrayLength + 1	exit 3	error	none
native ids renamed	critical	error	none
index removed	major (indexed)	valid	none

• mzValidate reports MS-Numpress as an error, though Numpress is standard mzML.
• Base64 broken across lines is valid mzML; mzdiff cannot see it until its reader reports it.

What it shows

• Every defect that changes Sage's matches or loses its spectra is major or critical in mzdiff, except base64 broken across lines, which mzdiff cannot see yet (Form).
• mzdiff reports several defects mzValidate passes as valid (a changed charge, a zeroed total ion current, a profile label, a parameter group), while mzValidate checks what a file must be on its own, such as index offsets and checksums, which mzdiff leaves to it. Run both.
• Some critical or major differences cost Sage nothing (the isolation target, collision energy, the filter string). Sage is one reader: DIA tools take the isolation target as the precursor, and spectrum predictors use the collision energy.
• Four severities were raised because of this check: a total ion current of zero, a dropped time unit, peaks in falling order, and precursors in another order (which Sage reads through the first precursor). The form: checks came from it too.