The Third Column: MS/MS fragment ion decharging in Mascot Server 2.7

In the Mascot Generic (mgf) peak list format, MS/MS fragment ions are typically defined as pairs of values. The first value is the m/z of the fragment ion peak and the second the intensity of the peak. However, a third value can also be supplied – the charge state of the fragment ion.

In Mascot 2.6 or earlier, the fragment ion charge state information if present would be ignored. This could present a problem if the MS/MS peaklists contained a large number of higher charge state fragment ions, as is often the case with peaklists from higher charge state precursors. In those cases, Mascot would, at best, fail to match the fragment ions. At worst you might get a false positive match to a 1+ or 2+ series. This issue is most pronounced in middle- and top-down experiments, where the higher charge state precursors are routinely selected.

This isn’t an issue if you processed your data using Mascot Distiller because you can get it to output a decharged peaklist (see this tip for more details). However, you may not have access to the raw data to do this, or, unlike Mascot Distiller, your peak picking software may not have the option to decharge peaklists. If this is the case, but you have peaklists with fragment charge state information (such as those generated from Bruker instruments) then Mascot 2.7 can automatically use this to decharge the fragment ion masses to MH+. This will often improve the score to a given peptide match because it allows for improved coverage of the peptide within the selected singly charged ions series used for scoring. The behaviour is controlled using a new configuration setting ‘DechargeFragmentPeaks’, which you can specify the maximum charge state to consider for decharging, or disable it altogether by setting the value to 0. This parameter can be embedded at the top of the peaklists file, in which case it will override the default setting of the Mascot server.

To see the effect of decharging on good quality MS/MS data, take a look at the following two matches for the same MS/MS peaklist from a middle-down dataset. The MS/MS spectrum was processed using Mascot Distiller so we have fragment ion charge states available and then searched using Mascot 2.7 with and without fragment charge states supplied in the peaklist:

• Peptide match with no decharging
• Peptide match with decharging enabled

Both matches are very good, but supplying the fragment charge state and allowing Mascot to de-charge them increases the score for the match from 100 to 124. This is because of the increased coverage of the peptide in the c, z+1 and z+2 ions series in the decharged result.

To examine the effect of this decharging on a more typical bottom up dataset, we took a single raw data file from a SILAC quantitation dataset generated using a Thermo QExactive from the PRIDE repository (PXD004607) and processed it using Mascot Distiller. The chosen dataset contains peak lists with precursors of charge states up to 5+, and roughly 1/3rd of the peaklists have precursor charge states or 3+ or higher. Therefore, we can expect to see many peaklists containing fragment ions with higher charge states and we would expect to get better results with decharging enabled.

Once processed, we exported the peak lists in the ‘comprehensive’ format. This is an extended mgf file with additional columns containing peak information, including fragment ion charge state. The peaklists were then searched using Mascot 2.7 with decharging enabled and disabled, and the number of significant PSMs in each of the search results was taken at the default significance threshold of 5% and also when adjusted for a 1% false discovery rate. Results are presented in Table 1 below.

	No decharging p < 0.05	Decharging by Mascot p < 0.05	No decharging 1% FDR	Decharging by Mascot 1% FDR
Significant PSMs	28691	30725	24029	28130

Table 1: Counts of significant PSMs when dataset was searched with and without fragment ion charge states present in the peaklists.

We found a significant increase in the number of peptide matches when the peaklists contain fragment ion charge states and are decharged by Mascot. At the default 5% significance threshold, enabling fragment decharging in Mascot 2.7 increased the number of significant PSMs by approximately 7%. At a 1% FDR, the results are even more pronounced and we found approximately 17% more matches.

Whether or not using fragment ion charge state information is a good idea will very much depend on your data. You need good quality MS/MS data to be confident that your peak picking software is providing accurate charge state information. With poor quality MS/MS data, charge state detection is likely to be unreliable, and you’ll probably get worse results. So we recommend searching with and without decharging enabled and comparing the results. Also, if you are routinely searching decharged peaklists then you should add instrument definitions to your Mascot server which only look for singly charged ions series, because the DechargeFragmentPeaks setting does not override the instrument definition.

Keywords: fragmentation, MGF, peak picking

Comments are closed.