Tuesday, October 19, 2021

It's over. SOMASCAN crushes mass spec proteomics!

 


We all knew this was coming, right? Well, here it is.


Doesn't sound like the title that is the end of mass spectrometry? Well, they understated it. Pull down supplemental table 2. 

Turns out that SOMASCAN is WAY better than an Orbitrap Elite running TMT 6-plex reagent on human plasma. You heard that right. 2013's best technology ran well (the data was actually acquired in 2013) on one of the most difficult matrices, can't quantify as many proteins as SOMASCAN.

 How's the quan compare? The SOMASCAN is still better according to all this expertly crafted manuscript.


Wait. It looks like I wrote this, but I know I didn't because I can't follow any of this math. 

Okay, so the LCMS data is from an instrument that you can pick up at ReUzeit with a warranty for around $50k, so what? (I still love the Orbitrap Elite and would definitely take a free one if you have one you'd like to get rid of. Awesome frustrating to calibrate monsters that they are). 

They also correlate their SOMASCAN data with GWAS, which is a condemnation completely of it's own. Look, I'm not saying that SOMASCAN doesn't work, but if I was gonna try to prove a technology worked, I probably wouldn't go digging for the oldest proteomics files on MASSIVE and try to correlate my protein measurements with some 23andme data. 

Update 4/2/2022 -- Ummm....uhoh....okay, I recently learned that several of the top hits for SOMASCAN link to this blog despite the millions of investor dollars the company dumps into buying ad space. 
And that makes me afraid that a lot of traffic coming here is not going to catch the problems mentioned above. 

For a more thorough breakdown, I recommend you check out this post

However, just to summarize this post, I really put that title up to annoy a lot of other scientists. I honestly tell graduate students in our lab that they have my permission to ignore anything done in proteomics before 2017. There are exceptions of course, there was some great work, but the instruments were so slow that thorough experimental designs were very difficult to pull off. A lot of the problems in proteomics can be attributed to the lack of technical and biological replicates, but each sample today can take you between 10 minutes and a couple of hours. When I first moved into proteomics, the same coverage (at lower quality) would take days or weeks. If I were to compare my preferred hardware to another lab's hardware, I wouldn't be making a very good point if I put our newest instrument to their oldest, and I feel like this is what happened here. 

No comments:

Post a Comment