Depends on the instrument, do you have a standard digest (ie HeLa) or something to benchmark the condition of your MS ?

From what your saying it seems like trouble at the fragmentation level but it could also be sample related so I would start there. Also what is normal for your TIC, are the counts and intensity units where you expect them to be ?

I second Tanveer's suggestion of a standard. We routinely use a commercial albumin digest, e.g., from Agilent; running an aliquot before starting a new series of samples, and any time there seems to be a problem. These digests are usually stable and - if you follow the instructions (add the correct amount of the suggested solvent) - are semi-quantitative, so you can get an idea both of the sensitivity and the fragmentation effectiveness.

Also, you mention that the total-ion current looks normal, but what about the actual CID spectra?

I agree, a standard would have been useful. Unfortunately there was none run close to this sample.

The TIC intensities (as shown by Xcalibur) have their maxima in the E9 range and them mean in the E8, which is about what I would expect (QExactive). The MS2-TIC (as given by MaxQuant) is in the E5 to E6 range, again as expected. In your opinion, is this the variable that should have most influence on the score? I find surprisingly little correlation between them.

What would be the appropriate quantity / quantities to judg the fragment spectra quality?

It could very well be that it is not a MS issue at all but something else entirely (LC, sample prep, column issues). If you're entirely sure that it is an MS issue, I would then suggest tuning and evaluating the q-exactive using their cal mix. Based on the evaluation, calibrate the parameters that need to be fixed. It's hard to judge what is the proper issue since there isn't some sort of standard.

As a start, I would recommend you keep a HeLa digest as a standard, I would suggest that over Albumin since HeLa would give you a better readout of Q-exactive performance due to it being a complex sample. Running a standard HeLa will also give you benchmark to base the instrument performance on since you will know what kind of numbers (peptides and proteins) to expect. Second, when running the standard I would recommend running low amounts (300-500 ng of HeLa tryptic digest on column) with maybe a 120 minute gradient, loading higher amounts is more likely mask performance issues due to overabundance. You should expect to see around ~2700 - 3000+ proteins and around 17-20k peptides of HeLa on a decent performing Q-Exactive with the above mentioned parameters. Lastly, get to know the instrument, and learn to look through parts of the gradient and look at the MS and MS/MS spectrum, see if it looks normal, look for strange background peaks (ie peg or some other background ions can interfere with the MS due to their over abundance filling the c-trap), also scroll through using arrow keys and see if it is performing Top N during the busy parts of the gradient like you expect, or if it's running out of candidates to fragment etc...

Ok, your advice on the standard sounds very helpful. I`ve had a general look at the gradient and spectra as you said, but was wondering if there was more that could be done.

I also would by no means rule out that the problem is in the sample rather than the instrument. But then there is still the question how to find out what is wrong with it.

Hmm without knowing much about the sample, you said your TIC looks normal so it would seem MS is fine. It could be an ion transfer issue/the quad issue where the calibration etc.. is off which is preventing proper isolation of the peptides for sequencing and thus you're getting less/wrong ions in the C-trap. Your best bet is to do the tune and evaluation in Xcalibar, the evaluation results would tell you if it's an MS issue and you can calibrate accordingly.

Hello Michael,

If you could rather exclude problems with MS, try to perform an error tolerant search with mascot. Have you done any chemical treatment (like isotope labeling) with your sample? An error tolerant search can help you find modifications you haven´t considered so far. Some time ago I had the same problems with samples labeled with isotope labels and found conclusions what went wrong during my sample preparation (labeling of N-termini in addition to labeled lysine residues).

Sounds like a good idea!

All good suggestions so far. I would suggest two additional things: a) run a MASCOT search with a relatively wide mass tolerance both in MS (say, 50 ppm) and MS/MS (say, 0.2 Da) to see if your calibration is for some reason way off. Although it would surprise me in a Q Exactive, this can happen. b) Use RawMeat to judge the number of MS and MS/MS scans. Compare the ratio to your previous, more successful experiments. c) when the instrument is next calibrated, run the Quad Transmission Endurance Test using the ESI source. If a Q Exactive gets dirty it would show something like this - it will still produce a nice TIC and many MS/MS, but the quality and therefore conversion rate of the MS/MS to sequence will no longer work out. The Test should give a value of 0.7 or better (out of 1); if not, the instrument needs a full quadrupole clean. And d) - yes, you will want to implement standards more in the future. Good luck!

Well, I am surprised that you could assigned more than 50% MS/MS spectra for DDA experiments before. Anyways, first check if the calibration is ok on your instrument. Take your top scored assignments from that run and look at the distribution over the theoretical and experimental mass difference. Secondly, your sample is too complex and most of MS/MS spectra are mixed: increase LC gradient time, decrease isolation window to 2 Da. Finally, the sensitivity of the instrument dropped for a number of mentioned (and many more) reasons (ESI, dirty quadrupoles, etc.)

Good ideas so far.. I usually go in this point-by-point scheme.. ;)

What is the sample? How was the extract prepared? (blocked cysteine using what? How many proteins do you identify?

1) for Q-Ex we do also see a similar assignment rate in a complex (e.g. human sample > 50%)

2) sample prep artifacts could be an issue and it can easily be identified by either Mascot-ErrorTolerant search or ProteinPilot search (open-mod-search with xxx modifications simultaneously)

3) if this is a low complex sample and a stretched gradient, this would also be ok ;) (assignments ~ 20%)

4) mass accuracy shifts of MS1 or MS2 can be tested by opening the window for searching (use something else than MaxQuant to do this)

5) try also another search engine ;)

For item 4 in Jonas's message you can use our library Pyteomics if you do scripts in Python: http://pythonhosted.org/pyteomics/

Out of curiosity: did you try the same experiment again and obtained the desired result (e.g. >50% assigned spectra)? Or tried it before?

What I mean is: was it the first time you tried this experiment? because in this case, it would be helpful to know what kind if sample type, sample prep protocol and LC conditions were used.

If it is a calibration problem... well you should see similar issues also in all the following samples. If this is not the case, then, as Tanveer suggested, the problem probably lies somewhere upstream of the ms.

Other possibilities: in your identification program, you did not allow for a post-translational modification that actually exists in a good number of the proteins in your sample. Or, in sample preparation, there was proteolytic digestion. Database matching software for MS-MS based identification only matches to peptides (usually only tryptic peptides unless you used a program that allows a different option) of specific size in your database, taking into consideration only the post-translational modifications that you set at the start. You could complement this identification process by using a program based on de novo sequencing. Peaks offers a trial version (http://www.bioinfor.com/peaks/features/peaksdb.html).

Thanks a lot for all the suggestions, this has been immensely helpful! The error tolerant search points towards a problem with the sample preparation which explains the low identification rate. I`ll include your suggestions in my further trouble shooting strategy.