Tuesday, October 20, 2015
TMT normalization based on variation is real important in clinical studies!
Any time we quantify anything we need to determine what is the priority for the measurement. Is it absolute quantification or is it lots of data points. This is always consistent, whether you're weighing out masses for buffers or doing global -omics quan.
We're absolutely no different. On Monday I worked with an awesome team developing an absolute quantification assay. One biomarker with heavy labeled spike-ins. One data point with precision and accuracy. But ONE measurment! Can we get that level of perfection with 16,000 measurements? Yeah...if you've got 10 years...(as technology improves...maybe 4 years?). And this will be measurement where we've reduced the level of technical variation.
Okay. So what happens when we look at something from a global level where the technical variation is high, but also the biological variation? I just learned that you can gain a lot of insight from this output by actually looking at the variation itself!
In this study in PLOS one, Evenlyne Maes et al., take a look at TMT labeled clinical samples and use statistical tools to normalize the samples using measurement variation. And all the sudden the data appears a lot more meaningful.
Variance normalization isn't a new concept in reporter ion based quantification. IsobariQ is a software a friend of mine has evaluated and really likes that was first showing off its algorithm for variance normalization in iTRAQ at HUPO in 2011. And there is an R package out there that does something similar and, for the life of me, I can never ever remember what it is called. Someone just asked me about it last week and I still haven't come up with the name.
What is cool about this paper is that we actually see the normalization working in clinical samples. Maybe someone else has done this, but I haven't seen it. If you show that something works well in your cool cancer cells that are grown at exactly 37C and get exactly 30% N2, or whatever, that is one thing (and awesome! and I'm not going to put it down ever. Bravo! I can't keep cells alive at all!)
But if you show me that your technique can extract more meaningful biological data out of people who walked into a clinic -- people who may have eaten 10 minutes ago, or 6 hours, or who have immunity to this virus or that or have any amount of the crazy differences each one of us walk around with every day? Thats gonna make me think we should take a look at what you did!