Friday, July 27, 2018

PPM or Da -- what to use and when in data processing?

A reader sent me in a suggestion to write this post -- coincidentally, it has been at the forefront of my mind lately cause I'm studying 2 PTMs that differ by less than 0.2 Da -- sound like fun? It TOTALLY IS.

Regardless of what search engine you're using you probably have at least 2 choices for mass tolerance for your engine. There are some smart new software packages that look at your data and select this for you now, but they aren't the norm yet.

Let's look at an example first -- I just updated SearchGUI/PeptideShaker and wanted to take a look at them anyway (best free de novo software ever!!)

Every engine has this somewhere.

Simple answer is this --- for ion traps and TOFs use Da and for FTICR and Orbitrap always use PPM.

End of post!

Not enough? Oh -- and in my videos and screenshots I normally use 10ppm MS1 and 0.02 Da MS/MS for the Orbitrap stuff? Okay -- that's because I'm lazy -- and that was before I started studying some PTMs that suck.

In an ion trap or a quadrupole or a TOF instrument the mass accuracy doesn't change as a function of the size of the ions you are looking at. An ion trap is accurate to the first decimal place as a function of the isolation and ejection speed. An ion trap running really slow (Enhanced mode or super slow mode or Zoom depending on what vendor you're using) is generally accurate to 0.2 and one running fast to 0.6 and one running super fast is probably closer to 0.9.  It doesn't matter if that ion is m/z of 50 or m/z of 1,500 the accuracy is pretty much set.

If you're on a hybrid instrument where you're doing MS1 with high res (like an Orbi or FTICR) and fragments in the ion trap -- these mass cutoffs should apply only to the fragment ions.

Similarly, a really good TOF is accurate to the second decimal place (I'm generalizing here) and it also doesn't care what that mass of the thing it's looking at is -- it's probably +/- 0.05 whether it is at m/z of 50 or m/z of 15,000. In general, the mass doesn't play in. Find the mass accuracy that is within the calibration range of the instrument and run with it.

Orbitraps (and -- I'm pretty sure, but I have VERY limited experience -- FTICRs) are different in this regard. The mass accuracy and resolution of an Orbitrap are affected by the size of the thing they are looking at. This is why we should really always use PPM. For global stuff where we're going to have filters and FDR and stuff -- it probably isn't as important and that is why we can use 0.02 Da for MS/MS tolerance on a QE -- until something weird comes up.

Right after you calibrate a Q Exactive the mass accuracy is going to be within 3ppm -- this is an average. If you run it nonstop for 5 days and then check the calibration it's probably going to have drifted out a few PPM more. I expect to get to 6ppm by the end of the week (MS1). Despite what I've seen claimed a few times -- lower resolution MS/MS scans do have slightly lower mass accuracy (I have proof now, courtesy of MetaMorpheus's recalibration function) it's not much, but it is a little and if you want a boring explanation, I have a very poor animation I could share.

Let's assume 10ppm because the math is easier.
On a 400 m/z ion 10ppm is +/- 0.004 --> that's crazy accurate!
On a 4,000 m/z ion 10ppm is +/-0.04 --> that's 10 times worse. If you have your mass accuracy at 0.02 Da on fragmentation of big ions like LysC digests or glycopeptides, you might miss valuable assignments on the largest ions -- and pick up noise you don't need on the smaller ions. Using PPMs that match the accuracy of your instrument help you avoid that stuff.

I've often ran into workflows that use 0.05 Da mass accuracy on a Q Exactive. This is almost always WAY too loose. On high mass fragment ions, this gives you enough wiggle room to allow the search engine to pick fragment ion M+1 isotopes for fragment confirmation. It'll look like this ---

This is generally not a good thing....and if you're doing something like heavy H+ labeling, this could be disastrous....  (Heavy carbons are evenly distributed across all the carbons in a peptide, there is no reason I can think of where selection of the fragment heavy isotope is ever a good thing)

Was that too many words? Probably. Hope it's helpful -- ping me if it isn't! And I'll tag this to the NewBies section later!


  1. I would love to know more about this explanation of lower mass accuracy with decreased resolution, say going from 30k to 15k on the MS2 (OT), especially since the opposite direction doesn't seem true (going from 30k to 60k).

  2. Do you happen to know what mass settings are ideal for a standard linear ion trap? (LTQ-XL)?