Nature methods - drift time based CE for improved fragmentation

[tsbatth]

http://www.nature.com/nmeth/journal/vaop/ncurrent/full/nmeth.2767.html

Cool paper on using Waters synapse to get fairly decent coverage. Good to see greater drift time applicability for this kind of stuff and alternative MS technologies challenging the Orbi in the numbers game. Haven't read through it thoroughly maybe some of you have and can share your insights .

[Zhang]

It is indeed an interesting paper, but the with an unfair comparison to Orbitrap-DDA methods. The Q-Exactive data were acquired from inappropriate settings. Therefore the quality was so poor to compare with their IMS-type of DIA on a TOF instrument.
The conclusion of "DIA can achieve higher levels of performance compared to DDA-based workflows, in terms of sensitivity, reproducibility, number of identified proteins and rate of peptide identification." is generally not true.

Here is a reference of higher coverage: Rapid and deep human proteome analysis by single-dimension shotgun proteomics.
Here is a reference of faster identification: The one hour yeast proteome

I don't know why the editor and reviewers of Nature Methods agreed on the statements.

[tsbatth]

I would have to somewhat disagree in regards poor quality of comparison. You're correct direct comparison might not be possible since a lot of times labs reporting these large numbers are usually the only ones that can do it consistently and each lab seems to have their specialty. Regardless, I think their statements holds for a few reasons. Fundamentally it's totally different than anything else recently and the fact they also release a DIA search engine is pretty impressive. In regards to comparing with the long gradient MCP paper, you have to keep in mind the drift time paper is using 10 times less starting material (300 ng vs 3ug) and a column that is half the length (25 cm vs 50) than in the MCP paper. I do have to say it's a bit curious that they tried at 400 ng and concluded that higher amounts didn't lead to more hits. But if you look at the supplemental information, even with lower loading amounts and half column length, they are getting peptide numbers higher than the long gradient paper in MCP. Even then, I agree the comparison isn't fair since the MCP paper used A375 instead of a common HeLa cell line for their study, which is a bit odd to me since HeLa seems to be the standard in most number papers these days. But it very well could be that IMS-TOF is more sensitive with lower starting material but Orbitrap instruments are better at handling more complexity at higher loading amounts.

The one hour yeast proteome, you cant compare either since it's different organism to begin with. Although the numbers with the fusion are impressive, I think the statements in that paper, especially the title is also a bit odd because if you include total LC run time with the loading time it amounts to something like 1.7 hours (and host of other issues I could go into).

[Zhang]

The reason why they load 300ng is further increase on-column loading will decrease the performance, not like the Orbi ones.
However, in Orbi experiments, with the 60 ms max-injection-time they almost never let the instrument to accumulate enough ions into collision cell. Besides, they used so wide a scan range in MS1, 300-1750. I mean seriously, how many peptides can be identified under 400.0 Th or over 1400Th? At the same time, they used two lock masses in every scan, MS1 and MS2, which significantly slowed down the data acquisition.

In my opinion, higher injection but lower identification means, when complexity goes up DIA will fail. However, DIA was designed for high complexity in handles multiplexed MS/MS. And the number game is also questionable, because what you search from DDA is the complete proteome sequence database, but for DIA it's only an MS/MS database.


I agree the one hour yeast is as well playing number games, and the different organism could not be directly compared, so is DIA vs DDA, and so is TOF vs Orbi. I was arguing that the statement in this DIA paper is misleading, trying to convince people that DIA is a superior method that beats DDA from every aspect. Shame on the reviewers, they agreed!

[Craig]

And the number game is also questionable, because what you search from DDA is the complete proteome sequence database, but for DIA it's only an MS/MS database.

Just to clarify, in this case they actually are searching a complete proteome sequence database for their DIA, not MS/MS spectral libraries. They use ProteinLynx GlobalSERVER (PLGS) from Waters which is a search algorithm for MS[SUP]E[/SUP] data.

[Infinity]

More on the DIA vs. DDA... Look at another paper in the same journal:
http://www.ncbi.nlm.nih.gov/pubmed/24162925
They tried to show advantages of DIA approach which is for me are not clear. First consider that prior to DIA you have to perform similar experiment in DDA mode just to get MS/MS library to use in DIA (!!!). And despite having results of both DIA and DDA approaches they haven't tried to compare them (which probably means that their comparison was not in favor of DIA).

[Zhang]

Maybe I was wrong on that. Unfortunatly, as a dry-lab student, I have my limited knowledge on Oribtrap, know little about Waters instruments and the data analysis.

[Zhang]

Indeed, in the same issue, just several pages before, you can find another article using SWATH (DIA). Mate, that's dominating!

More on the DIA vs. DDA... Look at another paper in the same journal:
http://www.ncbi.nlm.nih.gov/pubmed/24162925
They tried to show advantages of DIA approach which is for me are not clear. First consider that prior to DIA you have to perform similar experiment in DDA mode just to get MS/MS library to use in DIA (!!!). And despite having results of both DIA and DDA approaches they haven't tried to compare them (which probably means that their comparison was not in favor of DIA).

[m/z]

Maybe this paper that also uses ion mobility is relevant for this debate:

Advancing the High Throughput Identification of Disease Specific Protein Signatures Using Multiplexed Ion Mobility Spectrometry Mol Cell Proteomics January 8, 2014

, the purpose is:
"In an initial evaluation of the new LC-IMS-MS platform, its performance was compared to an LC-MS platform (comprised of a commercially available LTQ Orbitrap Velos)."
, and they conclude:
"... Overall, the LC-IMS-MS platform detected peptides at concentrations ~100× lower than the LC-MS platform with a linear correlation to concentrations"