How to design annotation files for CPTAC experiment ?

[winnieWei123456]

Hello, when using CPATC data in MSstatsTMT, I would like to inquire about how to fill in the annotation file.

The experimental design of CPTAC is very complex. For example, in this screenshot, there are 15 mixtures, each with 4 fractions. The 11 channels of each mixture are from different samples, which have different conditions, and each channel of each mixture may be a different sample. In this case, I would like to know whether TechRepMixture should all be set to 1? Because it seems that there are no technical replicates (TechRepMixture). How should Bioreplicate be set?

In addition, I would also like to ask how to combine different fractions of the same mixture. Can MSstatsTMT automatically merge them, or do they need to be manually merged? I am not sure how to analyze this experimental design.

Looking forward to your reply. Thanks a lot

[tonywu1999]

MSstatsTMT_annotation.csv

Hi,

I attached an example of what the annotation would look like. We will be creating a vignette with the CPTAC dataset soon. To answer your questions:

In this case, I would like to know whether TechRepMixture should all be set to 1? Because it seems that there are no technical replicates (TechRepMixture).

TechRepMixture can be all set to 1 in this case.

How should Bioreplicate be set?

Each sample should have its own BioReplicate ID. Using CPT0079300001_N BioReplicate ID as an example, you'll see that this ID is shared across multiple fractions (e.g. 1,2,3) and runs of the same mixture. It also is assigned to only one channel (128N), and one mixture (plex16).

In addition, I would also like to ask how to combine different fractions of the same mixture. Can MSstatsTMT automatically merge them, or do they need to be manually merged? I am not sure how to analyze this experimental design.

MSstatsTMT will merge different fractions of the same mixture automatically ("If a peptide ion was present in multiple fractions, it was only kept in the fraction where it had the highest mean intensity" - Huang 2020 et al.)

[winnieWei123456]

Got it! Thanks a lot. It really helps!
By the way, I would like to bother you with some additional questions. When I checked the quantification results generated by MSstatsTMT,I noticed that some mixtures retained the 126 channel, while others did not. In cases where the 126 channel was retained, most quantitative results for proteins in these mixtures showed "NA" values. Could you explain what might cause this? Is it normal for the final results to exclude the 126 channel, given that it serves as a reference channel?

[winnieWei123456]

Here's another question. Please forgive me as I have no experience in TMT data analysis. Does MSstatsTMT automatically account for batch effects in its analysis pipeline? I mean process the data to allow comparision between different mixtures. I think it has done this for the users.

[tonywu1999]

@winnieWei123456

Apologies for the delay in response. We generally check the MSstats google groups thread for inquiries. If you have more questions, could you follow up there?

By the way, I would like to bother you with some additional questions. When I checked the quantification results generated by MSstatsTMT,I noticed that some mixtures retained the 126 channel, while others did not. In cases where the 126 channel was retained, most quantitative results for proteins in these mixtures showed "NA" values. Could you explain what might cause this? Is it normal for the final results to exclude the 126 channel, given that it serves as a reference channel?

Thanks for your question — it's a good one!

In my experience working with a subset of the CPTAC TMT dataset, I haven’t seen the 126 channel consistently showing mostly NA values. Also, in some experiments, the 126 channel isn’t always used as the reference channel, so this behavior isn’t necessarily expected or "normal."

That said, there is a common pattern in TMT data that’s worth knowing about:
Within a single TMT mixture (i.e., within one mass spec run), if a protein is missing (NA) in one channel, it's often missing in all the other channels too. And if it's present in one channel, it's usually present across all the channels in that mixture. This is because the presence or absence of a protein's signal is usually tied to whether its peptides were successfully identified in that particular run.

In contrast, across different mixtures (since each is run separately on the mass spec), the same protein might be detected in some mixtures and missing in others. That’s often due to the instrument detecting different sets of peptides in different runs — this is a typical source of missing data in TMT experiments.

So in your case, the missing values in the 126 channel might be happening because the protein wasn’t identified in that mixture at all — not because there’s something wrong with the 126 channel specifically.

Does MSstatsTMT automatically account for batch effects in its analysis pipeline? I mean process the data to allow comparision between different mixtures. I think it has done this for the users.

Yes, MSstatsTMT does account for batch effects when comparing across different mixtures.

Each TMT mixture is run separately on the mass spectrometer, so there can be differences between mixtures just due to the instrument run and mixtures themselves. To correct for this, MSstatsTMT uses the reference channel in each mixture to bring everything onto the same scale, making the mixtures more comparable.

On top of that, MSstatsTMT also takes into account the fact that samples within the same mixture tend to behave more similarly to each other than to samples from other mixtures. It does this by including "mixture" as a random effect in the statistical model. That helps improve accuracy when estimating changes between conditions.

All of these steps are built into the MSstatsTMT pipeline automatically, so users don’t need to do them manually (all that is needed from the user is to annotate which sample/run belongs to which mixture and which channel is the reference channel for each run). You can read more about this in Table 1 of the MSstatsTMT paper.