Intact Mass includes a feature not found in other charge deconvolution software: a way to use known or likely mass differences in charge deconvolution. For example, 162 Da is a likely mass difference for glycosylated proteins such as monoclonal antibodies, and 22 Da is a likely mass difference for almost any sample with salt adducts. Intact Mass can accept up to three known mass deltas in its advanced configuration parameters. Known mass deltas are used as “hints” to help determine the charge of m/z peaks.
For example, if two m/z peaks differ by 162 / 7, then the algorithm bumps up the probability that the two peaks contain some charge-7 ions. Notice that known mass deltas act only upon charge probabilities, so they do not change the m/z spectrum at all, and change the deconvoluted m spectrum only indirectly. The default weighting of known mass deltas is light, so that charge assignments from known mass deltas are easily overruled by isotope peak spacing or charge series. With extremely noisy data, a known mass delta that is not actually in the sample may cause small artifact peaks on either side of a true peak, but this cannot occur unless there is m/z intensity at the necessary m/z offsets that is not removed by the baselining operation.
Figure 1. Without a known mass delta, the deconvoluted mass spectrum contains artifact peaks with masses below 27 kDa or above 33 kDa.
Figure 1 shows a deconvolution of a heavily glycosylated protein, recombinant human Epo, using the default parameters (except for m/z and m ranges). The major peaks in the deconvolution highlighted by colored dots are correct, but all the peaks below 27,000 Da or above 33,000 Da are artifacts caused by chance ratio relations among m/z peaks. With only three charge states and over 100 proteoforms in this native-MS spectrum, ratio coincidences are almost inevitable.
Figure 2. A single known mass delta of 657 Da, the average-isotope mass of HexNAc-Hex-NeuAc, removes almost all artifact peaks. One more mass delta, either 162 for Hex or 366 Da for HexNAc-Hex, cleans up the remaining small artifacts.
Figure 2 shows how to correct the problem with the use of an advanced parameter command, “KnownMassDelta”, to specify a likely mass difference for molecules in the sample.
Figure 3. Multimers are especially challenging for charge deconvolution, because distinct masses map to the same m/z peaks. The intact molecule in the spectrum above is 63,201-Da cholera toxin pentamer with five GM1 oligosaccharide ligands (999 Da), with a z = 11+ peak at m/z 5730. The intact molecule is fragmented to produce tetramer with 3 – 5 ligands at 49 – 52 kDa, trimer with 2 – 4 ligands at 36 – 39 kDa, dimer with 1 – 3 ligands at 24 – 27 kDa, and monomer with 0 – 2 ligands at 11 – 14 kDa. Notice that the m/z peaks with unequal numbers of proteins and ligands (colored dots) are not shared by multiple m peaks. Charge assignment is easier for unshared m/z peaks, and propagation of expected intensities via known mass deltas helps to divide shared m/z peaks.
Figure 3 shows the use of multiple known mass deltas—up to 3 are allowed—to solve a more difficult case.
Figure 4. For spectra with an evenly spaced sequence of mass deltas, as seen in the spectrum of an empty DMPC lipid nanodisk above, the parameter “CombFilter=5” directs the software to consider peaks differing by ± 677.5, ± 2 × 677.5, … , ± 5 × 677.5. If more than one known mass delta is specified as in Figure 3, a “CombFilter” value of 1 will be used with all mass deltas. The m/z spectrum above is especially difficult, because charge states overlap and disagree on relative peak heights.
Figure 4 shows the use of another command “CombFilter” to obtain a reasonable deconvolution of complicated spectrum with overlapping charge states.