Tuesday, May 31, 2016

Can't figure out your quantitative proteomics? Try a triangle!!!!

Either my stewardess really liked me and was extremely liberal with my dinner of scotch on my way to San Antonio, or this paper is completely brilliant. Considering my heritage, I'm leaning toward the latter.

What am I rambling about now? This fantastic new paper from Fernando García-Marqués et al., that you can find here.

Every single time we try to pigeonhole biology we're disappointed, because we completely underestimated it. A recent approach for the -omics community is something like this: (take a deep breath).

Forget everything that we think we know. And just look at what is statistically changing. Seriously. Throw out the gene ontology. Throw out the annotations. Start with what is statistically 
significant and throw the rest out. 


See if anything that we have figured out about the biology makes sense in this context. This is a tough one, I know. But it appears to be working. Think about it. How many times have we said "I've got this one. This is a checkpoint protein. So and so  proved it by western blots in '87 and about a thousand papers show that its a checkpoint protein. And then you find out that if you get a proteoform of it that is shortened 43 amino acids and is methylated that is has NOTHING to do with cell cycle checkpoints? Bonus points if you know what I'm talking about. Seriously. Ontology is awesome, but if we think we've got a protein down to "this is what it does" we are often surprised by something new it does. Its energetically favorable for evolution to find a bunch of new things for proteins to do. Its a lot less favorable to evolve a bunch of new proteins. 
The triangle approach isn't quite as intense as this. (Seriously, we do know something about biology by now). But you could cluster it in this regard. 

They start with their algorithm:GIA (which you can download here): 

GIA is short for "generic interpretation algorithm" YEAY! Thats what I'm talking about!!

And they take this triangle approach. There are a bunch of equations that are mostly Greek that walk through how they came to their approach. When it comes to the math stuff, I'm not the most qualified guy, but this seems really elegant and logical.  Start out by taking a step to the side. Assume you don't know what you're looking at. Let the Systems Biology Triangle do the hard thinking. Then once its done its pairwise analysis of the things that are truly significant, then go to what we know regarding the biology. 

Okay. How do you test such a nutsy concept? You could pull some historic data out. And find what they were looking for but didn't find. That's a nice start. Or you could pull some complex and confounding biological models, run your own experiments, see that your results make sense when Panther came up with nuthin. And then you could go ahead and do the validation yourself -- on a time course where early in nothing that they tested could see any perturbations in the system. But hey could. And it successfully predicts what is happening earlier in the timecourse showing massively higher sensitivity(that's all I could come up with) to the shifts that are going on at the proteome level.

They walk away from this sweet idea after doing some level of validation on 3 (or 4?) separate experimental sets.  And the validation looks top notch.  If any of the authors of this paper happen to be in San Antonio this week, hit me with an email at: orsburn@vt.edu.  I'd seriously like to introduce you to my secret employer's downstream pathway experts.


  1. We were a little away from San Antonio last week (Spain, Europe). But if you have some data we can try some triangles and integrations ;). Thank you very much for your kind comments.

    1. Fernando, that's a dangerous offer. We do have some quantitative data sets (from my past life) sitting around that we couldn't make any sense out of. I'd be happy to dig through a hard drive one weekend and send you some stuff!