- Understanding the effects of mass tolerance settings in Byonic
Precursor and fragment mass tolerances are important input parameters for Byonic™, and appropriate setting of these parameters can help ensure the specificity and sensitivity of search results. There are four parameters to set in the Processing nodes tab:Precursor Mass Tolerance and Fragment Mass Tolerance under Instrument Parameters, and Precursor and charge assignments and Precursor isotope off by x under Spectrum Input Options. For both precursor and fragment ion scoring, Byonic takes mass errors into account, so that even with a fragment tolerance of 0.5 Da, a peak with 0.1 Da error (difference between the theoretical and observed m/z) scores more than a peak of the same intensity with 0.2 error. The user-set tolerance is the maximum error allowed in a match of an observed peak to a theoretical ion.
In the absence of information from Preview or previous Byonic searches, we generally set Precursor mass tolerance to 10 ppm for Orbitrap or QTOF MS1, and 1.5 Da for ion-trap MS1. Fragment mass tolerance, on the other hand, is generally set to 20 ppm for Orbitrap MS2, 30 ppm for QTOF MS2, and 0.5 Da for ion-trap MS2. We sometimes try a Fragment mass tolerance setting of 0.02 Da or 0.03 Da for Orbitrap or QTOF MS2 data that may have systematic errors, for example, data from multiple LC runs or multiple MALDI plates. With Byonic there is no need to set large fragment mass tolerances (for example, 1.5 Da) for ETD spectra in order to compensate for c−1 and z+1 ions, which are common in ETD spectra of low-charge (z=2+) precursors. Byonic predicts c−1 and z+1 ions, so you should set ETD fragment tolerances exactly the same way you would set CID fragment tolerances for the same mass analyzer.
We set Precursor and Charge assignments to Compute from MS1 for .raw, .mzML, and .mzXML data with high-resolution MS1. We then set Precursor isotope off by x to Too high or low (narrow). Byonic includes an algorithm for precursor monoisotopic mass determination that is about as accurate as “averagine” can be; its errors are infrequent and symmetrical around zero. For data without high-resolution MS1 (for example, a .raw file from an ion-trap instrument or an .mgf file from any instrument), we set Precursor and Charge assignments to Originally assigned, and Precursor isotope off by x to Too high (narrow) or Too high (wide). Too high (narrow) is better for bottom-up proteomics with few precursors over 3500 Da, and Too high (wide) is better for top-down proteomics, and sometimes for N-glycoproteomics, whenever many precursors exceed 4000 Da. Originally assigned monoisotopic masses (from XCalibur, Proteome Discoverer, Mascot Distiller, etc.) have asymmetric error patterns, with observed mass more often too high than too low, as an isotope peak is often mistaken for the monoisotopic peak.
Preview™ is a free software from Protein Metrics that allows the user to make more informed decisions of mass tolerances and other input parameters. As shown in Figure 1, Preview reports various mass error statistics; the important ones are “Median precursor accuracy” and “Median fragment accuracy”. We recommend setting the precursor tolerance to about 5 times the median accuracy, so in this case 5 ppm precursor. The fragment tolerance should also be set to about 5 times the median accuracy, but generally not below 10 ppm. Mass tolerances change Byonic scores, so we find it convenient to use round numbers as tolerances in order to make searches more canonical.
Figure 1: Preview reports precursor and fragment m/z errors both before and after recalibration. It then launches the PMI-Byonic App with tolerances set to about 5 times the “Before recalibration” median accuracies.
Preview writes parameters for a quadratic recalibration curve into Byonic’s .byparms file; the user can then edit the .byparms file to add or subtract modifications, change mass tolerances, precursor isotope off by x, and so forth. Recalibration will make little difference if m/z measurements are already well-calibrated as in Figure 1, but it can make a large difference on poorly calibrated data. Recalibration assumes that the same curve works for the entire data set, so it is not as effective with data with various error modes.
Figure 2: When launched from Preview, the PMI-Byonic App allows the user to recalibrate measurements and tighten tolerances.
We obtained data from blood plasma enriched for glycopeptides with wheat germ agglutinin, reduced with DTT, digested with trypsin, and alkylated with iodoacetamide. We searched the data for fully tryptic peptides with at most one missed cleavage, and used 10 ppm precursor tolerance for Orbitrap MS1 and 0.5 Da fragment tolerance for ETD ion-trap MS2. We used a small database containing about 200 abundant plasma proteins, including all the most abundant glycoproteins. We set glycan modifications with the database N-glycan 57 human plasma set to rare1, and added 7 common1 glycans individually to allow for two-glycan peptides containing one of these 7 along with one of the 57 rare1 glycans. This gives a slightly faster search than setting N-glycan 57 human plasma to be common1 or common2.
Byonic found about 35 glycoproteins in this small data set (2733 MS2 spectra), with multiple glycosylation sites and multiple glycans (up to about ~10 distinct compositions) per site. See Figure 3.
Figure 3: Glycopeptide search results as viewed in PMI-Byonic-Viewer. Notice the strong loss of NeuAc from the charge-reduced precursors in the displayed spectrum.