[Question] Recommended workflow for multi-batch TMT FDR control

[jgriss]

Hi,

First of all, thanks a lot for your great tools! They work perfectly and make my life a lot easier!

I have a theoretical question about how to perform the FDR control in a multi-batch TMT experiment. The setup is as follows:

• 3 TMT batches (each 16-plex), each with around 10 fractions
• 1 Bridge channel per batch

I am now wondering about how to correctly control the FDR. The workflow is clear for these steps

1. Search each RAW file (separately)
2. Run PeptideProphet on each search result to get probabilities (separately)

Question 1: ProteinProphet

Should ProteinProphet be run on the search results from all batches or only on the files from one batch?

Question 2: Filtering

In order to create the psm.tsv files I use the philosopher filter command. Same question as before: Should this be run on all pep.xml files simultaneously or split by batch?

Also, would you recommend to use the --sequential parameter for this setup?

Thanks a lot for your help!

Kind regards,
Johannes

[prvst]

Hi @jgriss.

Question 1: ProteinProphet
The ideal scenario is to use a combined protein inference for the FDR filtering. To achieve that, you need to run ProteinProphet on all experiments that you are planning to use in future comparison. If you want to include all three plexes in your study, then you run ProteinProphet on all interact.pep.xml files from all experiments.

Question 2: Filtering
The filter command needs to be executed once for each workspace. In this case, you have three experiments, then you need to execute three times. This will create a PSM table for each one.

I suggest that you run the program using the pipeline command. This option automates everything for you.

[jgriss]

Hi @prvst

Thanks a lot for the quick reply! This answers my question!

I'm aware of the pipeline command (which btw. is a really great feature!). As you may have guessed I currently need to run steps manually. We found some discrepancies with our default pipeline and I'm now in the process of tracking down the differences.

Just for me to get the theoretical part right:

• Creating the prot.xml creates a dataset wide protein model which is then used for a global protein FDR control
• The filter command calculates an experiment-level PSM FDR in addition to the protein FDR

Again, thanks for the great support and the help!

Kind regards,
Johannes

[prvst]

Thanks, I'm glad you find the program useful.
Yes, you are correct. The global protein inference allows you to have a single protein list that can be used later for comparing the output from each experiment. You then run the filter individually, to each experiment.