Sunday, April 6, 2014

Byonic + ptmRS

You probably saw this one coming.  Last week I think I added a total of 7 new nodes to PD 1.4 on this laptop (details on the other 4 are coming, don't worry!).  You know I'm biased, but I really believe that PD is rapidly becoming the most powerful and flexible proteomics package ever.  Yes, this is heresy.  I know the extreme love that a lot of you feel for the formerly great Elucidator.  But when you see everything that PD can currently do (and will soon be able to, hint hint...) I think we'll have something that will blow even that legendary software package away.

Obviously, I've been obsessing a little over these two in specific:  the Byonic node and ptmRS.  How well do they work together?  Are they redundant at all?  Turns out they are perfectly and amazingly complementary!

A big part of my weekend has been Byonic wild card searching RAW files that my collaborator and I have acquired over the last 2 years in the search for PTMs that might explain our low peptide identification rates in certain files.  (Yes, this is what I do for fun sometimes.  I blast death metal all weekend, drink nice cheap Riojas and search RAW files with cool programs y'all send me.  Yes, my new next door neighbors have already stopped talking to me...which is a shame cause they have a T.A.R.D.I.S bumper sticker and I figured we'd be friends...sometimes we just have to make sacrifices in the name of science!)

What I have found is a slight weakness in the Byonic node.  I emphasize slight because I haven't actually read the instructions.  I might be able to fix it by doing so.  I have found that the Byonic node sometimes can't exactly assign a new PTM that it has found and you occasionally get multiple peptide options for the PTM locations.  Easy example:

For this PSM, Byonic wild card search can't tell where the +57 mod is.  Presumably, this is carbamidomethylation on the cysteine, so it should be the second one.  The Byonic scores are virtually identical for the two options, so that isn't much help here.  Easy to solve this one, but this could be more daunting if I, say a phosphorylation on site 223 in p53 means one thing and a phosphorylation on 229 means another....

Again, reading the instructions and/or changing some of the many settings we can control in Byonic may immediately improve this, but I have ptmRS!!!!!

Question 1)  Are they compatible?  As you can see from the screenshot at the top, Proteome Discoverer certainly believes that they are.  When nodes aren't logical or compatible we can't draw arrows from one to the next.  Starting a workflow results in no errors.

Unfortunately, however, the workflows do not finish.  They end up crashing out.  But thats okay.  Because my laptop is FAST.

This is my current workflow of post translational modifications:

Byonic wild card search to find the PTMs that are present, followed by directly searching for those PTMs with a more conventional search engine:

SequestHT (for high-low data) MSAmanda (for high high data) ran with ptmRS.  While it would be useful to have the two programs running together, there is a big advantage for doing this way.

Confirmation!  If Byonic tells you you are looking at a mod in your data and MSAmanda confirms that observation AND ptmRS gives you site localization that makes a ton of sense, how are you doing anything but winning?

Just another advantage of developers really grabbing on to the Discoverer frame work.  We can not only get rid of apparent limitations in one algorithm, but also make strides toward turning it into an advantage!


  1. It depends on the fragmentation coverage of the peptide. Byonic can't assign if it can't identify the relevant fragment ions. when it does, the assignment is correct. that is common problem with weak spectra or large peptides.

  2. David,
    Thanks for the feedback! I know this is a common weakness in MS/MS shotgun data. I've been referring to it as "circumstantial evidence" PTM assignment. I love the phosphoRS, ptmRS output though and its amazing how complementary these nodes are, it just takes an additional run to statistically localize the PTMS and these data reports look amazing. PTMS I didn't know were there, verified by Sequest/Mascot AND localized!