Saturday, May 18, 2024

Global detection of human variants by proteomics - do we still need the transcript data at all?

 


This came out around New Year's and I hadn't gotten around to it. It is a very thoughtful analysis and on the surface you can flip through and think - well...shit....I guess we still do need those sequencing people. Particularly after seeing Figure 5A (above).

The transcriptomics does a better job of detecting all of these weird variants than the proteomics does. Way way bigger numbers. 

Oh yeah, this is the paper. Wow, did it have a fun time in peer review.....


And we know that, end of the day, transcriptomics data generation is a lot less expensive than we are. However --- it seems to me like transcriptomics data analysis is not doing the exact same thing. Depth is going up and so are false discoveries and the amount of super computer firepower to dig through all this extra depth. 

This group goes deeeeeeeep on the proteome. Multiple enzymes huge coverage. And here is the funny part, even when they should be able to see some of these variants the transcriptomics they don't. 

I suspect here is where the peer review went from weeks to months to many months -- what is the ground truth here? The authors suggest that the transcript variants simply aren't making it to stable proteins that can be detected. I'm sure some genomics guys as reviewers 2-4 were like - "....yeah, or maybe you and your whole field is not very smart....? We're so SMART, we sorta trademarked it! 

Either way, I think there are some fundamental questions here and some fundamental truths. 

Proteomics can be relatively inexpensive, but it can also be very very expensive. To get anywhere near the detection capabilities of transcriptomics right now you have to do the latter one. Data analysis of either -omics is not cheap. Well, it can be, if labor is free. Skilled labor that can determine the relative likelihood of a detected variant being true in either technology is not free and it probably won't be within my lifetime, regardless of how many NVIDIA boards you inefficiently sit close together and link together with a proprietary cable. You'll probably want someone to look at that and say "yup! ya skipped a big chunk of oligos right there and made a weird proteoform!" 

No comments:

Post a Comment