Tutorial 1: Confirming Polyisotopic Compounds in a Mass Spectrum
In this tutorial we will use IsoGlyph to confirm if a polyisotopic compound is present in a noisy experimental mass spectrum or not.
Preparing the data
Section titled “Preparing the data”The mass spectrum used in this tutorial can be downloaded here. It is a medium resolution spectrum that was produced by detecting a low concentration of HgCl2 with iodide reagent ion. The ion detectable in the spectrum is HgI3- and the signal-to-noise ratio is low.
The file HgI3_low_signal.tsv contains two tab-separated columns without a header. The first column corresponds
to m/z and the second column - to peak intensity. This spectrum has already been centroided. IsoGlyph also supports
profile spectra, which are supplied in the same tab-separated format, but the second column corresponds to raw signal
in that case. The easiest way to generate a .tsv file from your existing spectrum is to open it in Excel, copy
m/z and intensity columns, and paste them into a blank text file.
100.0025 6.183e0100.0349 2.436e1...Getting started with IsoGlyph
Section titled “Getting started with IsoGlyph”If you haven’t already, download the latest version of IsoGlyph for your platform. Extract the executable and move it to a convenient location. Open the program. Populate each field as directed below. Brief explanations of what each field does are provided here. For more detailed explanations of each input, click on the links in the list.
| Input | Description | Value |
|---|---|---|
| Path to spectrum | Absolute or relative path to the .tsv spectrum that you downloaded |
|
| Peak separation | Smallest allowed peak separation in the theoretical spectrum (amu) | 0.02 |
| Min. rel. intensity (theoretical) | Minimum allowed peak intensity in the theoretical spectrum relative to the tallest peak | 0.001 |
| Min. abs. intensity (experimental) | Minimum allowed peak intensity in the experimental spectrum | 0.01 |
| Tolerance | Maximum allowed m/z drift of peaks within a mass spectrum relative to each other (amu) | 0.1 |
| Find N candidates | Number of matched candidates to plot and display | 1 |
| Estimate and subtract background | Optionally, estimate and subtract local background noise near the matched spectrum | on |
| Search window width (tab) | Potential matches will only be searched for in the vicinity of the theoretical spectrum | |
| Width | Width of the search window, centered at the theoretical mass spectrum (amu) | 10 |
| Centroid spectrum (tab) | Spectrum is already centroided and each point will be treated as a peak | |
| Search mixture (tab) | Search will be performed either for a single compound or a mixture with known composition | |
| Compound | Chemical formula of the compound that will be searched | HgI3 |
| Fraction | Fraction of the compound in a mixture | 1 |
| Charge | Charge of the compound | 1 |
| Experimental plot scale | Switch between linear and log intensity scale on the plot with the full experimental spectrum | log |
Running IsoGlyph and interpreting the results
Section titled “Running IsoGlyph and interpreting the results”Positive match
Section titled “Positive match”With the example spectrum HgI3_low_signal.tsv downloaded and all IsoGlyph parameters set to values given in the
table above, click on “Run search” in the bottom left corner. The search should only take a few seconds and at the end,
some plots and text output will be displayed:
This search result suggests that HgI3 is present in the sample with high confidence. A user can consider several different IsoGlyph outputs as evidence of compound presence.
First, upon visual inspection of the top left plot, a user can confirm if the shape of the matched experimental spectrum is close to the theoretical spectrum. In this case, they overlap very well.
The next metric to consider is the fraction of matched peaks, which is shown in the table below the plot area. In this case, 7/7 peaks were matched, which is strong evidence for positive detection. However, if the fraction of matched peaks is low, but the denominator is high, the peak separation parameter might simply be too high for the resolution of the instrument. Increasing “peak separation (theoretical)” could merge theoretical peaks that are close to each other, reduce the total number of peaks that need to be matched, and potentially improve the quality of the match.
Another important metric is linearity of the parity plot (shown on the right). If the plot is visually linear and parity plot R2, which is displayed in the table below the plot area, is above 0.900, then this evidence supports a positive match.
Lastly, the signal-to-noise ratio needs to be considered. It is possible that all of the matched peaks are of comparable intensity as background noise, which would reduce the confidence in a positive match. There are two ways to estimate the local signal to noise ratio affecting the match. If “estimate and subtract background” was enabled, as in this example, then compare the order of magnitude of “inverse scaling factor” and “background”. Inverse scaling factor is the intensity of the tallest matched experimental peak. Background is a constant intensity that was subtracted from all matched experimental peaks. Ideally, inverse scaling factor should be much greater than background. If they are comparable, then it is possible that the matched peaks were noise and confidence is reduced. If “estimate and subtract background” was not enabled, then the area ratio parameter can be used as a metric of local background noise. If it is close to 1, the match was not affected by the background significantly.
Another metric to consider is the shift. If it is close to zero, then the matched experimental spectrum is at the same m/z as the theoretical spectrum. However, a large offset does not immediately disqualify the match, since a mass bias could be present in the measurement.
Negative or inconclusive match
Section titled “Negative or inconclusive match”Change the compound from HgI3 to HgBr3. Click on “Run search” again. This time, the top candidate does not appear to be a likely match. Upon visual inspection of the plot on the top left, the shape overlap of matched and theoretical spectra is poor. The parity plot does not look linear and R2 of 0.710 is well below 0.900. The matched spectrum is shifted to the left by 1.6 amu. That, by itself, does not serve as evidence against a match. However, HgI3 was previously matched in the same experimental spectrum with zero mass shift. Therefore, this measurement does not appear to have a systematic mass bias and a shift in the matched HgI3 spectrum suggests that it was probably matched to noise rather than true peaks. A search result like this cannot serve as evidence of presence of a compound in the spectrum. Unless tuning the search parameters can improve some of these metrics, the compound of interest was likely not present in the sample in a concentration above the detection limit.
