Dear Skyline Users,

Questions were flowing in during our third Skyline Tutorial Webinar -- and here are the answers! 

Q:  How many fragment ions are selected for quantification? Are smaller peptides at a disadvantage because they have fewer fragment ions?

Ans: When choosing the set of chromatograms from which peak areas are used for quantitative data for peptides in your experiment, there are two competing factors:

  1. Ion statistics - more ions generally mean better precision. This factor favors using any transition for which you have a measured ions.
  2. Probability of interference on any transition (i.e. selectivity) - the higher this probability is the more likely you will compromise your measurement by including more transitions, some of which may contain interference.

With a method that achieves the ultimate selectivity of 0% probability of interference (all signal with no noise), then you would include every ion for which you have a measurement. With a method that has much lower selectivity (e.g. chromatogram extraction from MS1 scans) you may well be better off using only the most intense ion, because then you expose your measurement to that probability only once, and on the ion that is likely to have the best signal to noise ratio. Our own prior studies in SRM indicated that using the 3-5 most intense fragment ions for quantification produced lower CVs than using the single most intense ion. A method as selective as PRM should see similar results.

Q:  Please explain how to find the reason for missing ID annotation in MS1 filtering data processing on Skyline. My proteins has very good hit on mascot but not showing ID for the peptides for this protein in Skyline with MS1 filtering experiment.

Ans: This was discussed in the first webinar. It is also mentioned in the MS1 Full-Scan Filtering tutorial. By far the most common case we see is with Mascot and the cause is that the MGF converter used to produce an input MGF file for Mascot did not include the necessary source-file and retention time information. By far the most common case of this is also caused by Proteome Discoverer 1.x. Though, we understand from Thermo that PD 2.0 fixes this problem. For more information on MGF conversion and understanding why ID annotations may be missing, please consult this tip:

https://skyline.gs.washington.edu/labkey/wiki/home/software/Skyline/page.view?name=mascot_missing_rt 

Q:  How many concurrent precursors can you have when running t-MS2 in a QExactive when using a scheduled method? For example, in a 2 h method, RT window of 2 min for a complex sample.

Ans:  We have estimated a maximum number of targeted concurrent peptides (overlapping monitoring windows) being measured in acceptable conditions around 60. The resulting number of peptides measured in the full analysis will strongly depend from the distribution of their elution time. An in-depth discussion around these aspects is included in the following publication:

Technical considerations for large-scale parallel reaction monitoring analysis. Gallien S, Bourmaud A, Kim SY, Domon B. J Proteomics. 2014 Apr 4;100:147-59. doi: 10.1016/j.jprot.2013.10.029. Epub 2013 Nov 4. (http://www.sciencedirect.com/science/article/pii/S1874391913005447)

Q:  How many peptides can we realistically co-isolate and co-fragment in a QExactive instrument?

Ans: The co-isolation of two peptides (typically pairs of “heavy” and ”light” peptides) is the standard mode used for multiplex PRM acquisition. However, we have already demonstrated successful PRM experiments based on the co-measurement of four peptides while keeping sufficient selectivity (as individual isolation windows remain at 1-2 Th). This is presented in the following publication:

Targeted proteomic quantification on quadrupole-orbitrap mass spectrometer. Gallien S, Duriez E, Crone C, Kellmann M, Moehring T, Domon B. Mol Cell Proteomics. 2012 Dec;11(12):1709-23. doi: 10.1074/mcp.O112.019802. Epub 2012 Sep 7. (http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3518128/

Q:  How do you set an ion extraction tolerance in Skyline, of say 10 ppm?

Ans: The extraction tolerance in Skyline is based on resolving power (rp), which is really the important metric for separating signal for quantification on a mass spectrometer. As Bruno showed, lower resolution can lead to the inability to separate two peaks that can be separated at higher resolution. You may have a mass spectrometer with 5ppm mass accuracy, but if it has 10,000 rp, then you should be extracting chromatograms with a 50-100ppm window. If you are using centroided data, then your centroiding algorithm still will not be able to separate peaks in the m/z dimension that are closer than this. It is just that this fact, may be hidden from you with centroided data.

There is a nice example of this at the end of the DIA tutorial:

https://skyline.gs.washington.edu/labkey/_webdav/home/software/Skyline/%40files/tutorials/DIA-2_6.pdf#page=39

Q:  Is it necessary to acquire MS1 spectra in PRM experiments using skyline?

Ans: It is not necessary, but it can be useful extra information to be able to extract precursor ions from MS1 scans. On ion trap instruments it may be worth considering that some amount of time will be spent on MS1 scanning for automatic gain control (estimating ion flow to the trap). On Thermo instruments, including an MS1 scan in your acquisition cycle can alleviate time that would be spent doing MS1 scanning for AGC, though AGC-only MS1 scanning may require less time overall.

If you do not acquire MS1 scans with your PRM data, then you should turn off the MS1 extraction parameters in the Transition Settings - Full-Scan tab, and not include "p" for precursor in the Transition Settings - Filter tab.

Q:  If one is interested in very low abundance PTMs, such as O-GlcNAc and phospho-Tyrosine, which method would be best to quantify? We are doing SWATH, DDA, but it seems that we should be doing PRM.

Ans: One key factor is how many peptides you need to measure. DIA/SWATH and DDA are very well suited for large numbers of targets. PRM and SRM will likely give you better results with very low abundance peptides in complex mixtures. However, the MacCoss lab has had success measuring low abundance transcription factors with DIA (though not really with DDA). Skyline should provide you all the tools you need to do your own comparisons of these methods for your needs.

Q:  As the cycle time in PRM is about 2 seconds, this limits the number of the peptides we can scan in the orbi if we have to scan each peptide individually at a given time. Can we set the orbi to scan more than one peptide at once?

Ans: You can, and this was alluded to in Bruno's presentation. With a Q Exactive type instrument, you can multiplex isolation, and then scan a combined set of isolations. An interesting example that was presented was co-isolation of the light and heavy precursors in an experiment that includes stable isotope labeled internal standards, which should roughly double the throughput of such an experiment, over isolating each separately.

Q:  In your tutorial, you ticked the box "Filter by peptides in the document" when building the library. If I add an additional peptide to the Skyline file, do I need to re-build the library or is it automatically done?

Ans: It is important to remember that the data I built the library from was acquired in PRM mode. That means the instrument was isolating very limited ranges of precursor m/z potentially over limited time ranges (if scheduling was used). The likelihood that those ranges happened to isolate other interesting peptides is relatively low, unless they were something like different phosphorylations of the same peptied.

The answer is "yes" you would need to rebuild the library to get other peptides into it, but the likelihood that the peptide search results include other interesting peptides is relatively low, even lower, if you are using a high performance quadrupole with 0.7 or lower precursor isolation.

Remember, this is not DDA or DIA.

Q:  Is there a Skyline control with code or command line? Can you write code in R for example to analyse data in skyline?

Ans: There is a way to run Skyline from a command line, using a program called SkylineRunner, which you can download (with documentation) from the Skyline install page. It is important to note, however, that it really just runs Skyline without an UI. It requires Skyline to be installed on the computer as well, and Skyline only runs on the Windows operating system. So, SkylineRunner may not be well suited to running on a Linux cluster for example, unless you can work out the details of running it in a Windows VM there.

Skyline does not support being run directly from within R, but it does have extensive support for exporting it results to tabular formatted reports, and these can be easily imported into R for analysis.

https://skyline.gs.washington.edu/labkey/wiki/home/software/Skyline/page.view?name=tutorial_custom_reports

This is one of Skyline's strengths. Quite a lot of R analysis has been done on data from Skyline. Several labs have even written R programs that can be run as External Tools from within Skyline.

https://skyline.gs.washington.edu/labkey/project/home/software/Skyline/tools/begin.view

You could certainly write a program that ran SkylineRunner to export a report from Skyline, and then ran R code to analyze the report. We are doing exactly this in the MacCoss lab for system suitibility analysis.

Q:  How does the sensitivity for PRM compare to SRM measurements in your experience? Can PRM on a Q executive measure with the same sensitivity as SRM on a high end tripleQuad instrument?

Ans:  The LOD/LOQ obtained with PRM experiments are in general similar or even better than those obtained by SRM. An extensive comparison of the sensitivity of the two methods is included in the following publication:

Selectivity of LC-MS/MS analysis: implication for proteomics experiments. Gallien S, Duriez E, Demeure K, Domon B. J Proteomics. 2013 Apr 9;81:148-58. doi: 10.1016/j.jprot.2012.11.005. Epub 2012 Nov 14. (http://www.sciencedirect.com/science/article/pii/S1874391912007452).

Q:  Regarding the manual setting of integration boundaries, is it more favourable to set them narrow or broad? Is it acceotable to use second derivative in transform to set integration boundaries? Is it possible to set individual integration boundaries for each selected transition?

Ans: With regard to "narrow or broad", you should note that Skyline performs background subtraction, which makes this question a lot less interesting, because narrow v. broad mostly changes the width at background level, which gets subtracted. Jarrett mentioned this in webinar 2. For a full explanation of how Skyline calculates peak areas see this tip:

https://skyline.gs.washington.edu/labkey/wiki/home/software/Skyline/page.view?name=tip_peak_calc

Skyline uses second derivative smoothing in its algorithm to determine the boundaries it uses for integration, but it then further uses Savitzky-Golay smoothing, and finally just looks at the raw data. It is not recommended that you perform peak boundary adjustment while viewing only a second derivative smoothing of the data, as this will likely cause you to choose inappropriate boundaries. Early versions of Skyline used just second derivative smoothing, until it became clear that this had a tendency to set integration boundaries too narrow.

It is possible, but not recommended, to set individual integration boundaries for each chromatogram/transition. Experience has shown that this increases CVs in the data by adding human variance, when compared with Skyline using the same boundaries for all transitions with background subtraction.

Q:  Is there any penalty in Skyline for using centroided MS2 data in PRM on a Q-Exactive instead of profile data?

Ans: One penalty is loss of information. Profile mode data is a more faithful representation of what the instrument measured, while centroiding relies on another algorithm to transform the profile data into centroided points. In terms of data quality, however, the penalty should be pretty low to non-existent if you use an extraction range based on resolving power. You could still end up including more interference or losing more signal because of centroiding that combines multiple peaks that cannot be resolved from each other, but this is expected to be more of a problem, if the centroided data is extracted using a mass-accuracy-based filter (e.g. 5ppm) instead of a resolution-based filter.

Q:  After exporting results, what program do you use to do statistical analysis when comparing the control of treated/disease samples (several biological replicates and several targeted proteins)?

Ans: We have been working closely with the Vitek lab ot integrate their MSstats tool with Skyline, and the next version of Skyline will contain statistics calculated inside Skyline, but derived from MSstats. We think this collaboration has benefited both groups and both Skyline and MSstats software. With Skyline 2.6 (the current version), you can use MSstats as an External Tool directly from within Skyline.

https://skyline.gs.washington.edu/labkey/skyts/home/software/Skyline/tools/details.view?name=MSstats

There are other tools, though. I have heard a tool called DiffProt works well with MS1 filtering data and another currently in review called MAP-DIA works well for DIA data.

Q:  Is there a way to export the retention times from the spectral library. i.e. in the spectral library explorer?

Ans: As it turns out, there is not. Interesting oversight that we will try to remedy in the future. Thanks for pointing it out.

Q:  Approximately how many peptides are you able to quantitate with PRM in a 1 hour method?

Ans: The number of peptides possibly measured in 1-h analysis is tightly associated with the experimental parameters (e.g., monitoring window in the time-scheduled acquisition, distribution of the elution times of the peptides, resolving power, fill time). In our hands, by using MS acquisition parameters providing good sensitivity and selectivity and a precise control of the peptide monitoring windows, 500-600 peptides distributed over the full elution space can be measured.