Compatability with DIA-NN 2.0

[kswovick]

Hi,

I was wondering if you had any plans to update the DIA-NN converter in your future plans. With the new parquet file format (and no report.tsv included), there is no longer a "FragmentQuantCorrected" column, resulting in the following error:
"Error in [.data.table(dn_input, , req_cols, with = FALSE) :
column(s) not found: [FragmentQuantCorrected]".

Thank you for this great tool.

[tonywu1999]

Hi,

Could you elaborate on how DIANN reports the fragment ion quantifications? Which column do they use and what's the format there? Sample data would help too.

Tony

[kswovick]

It appears as if there is no column for fragment ion quantifications in this version. I do wonder if it has something to do with the QuantUMS algorithm since it uses both MS1 and MS2 for quantification. I'll try re-searching some data over the weekend in version 2.0 but using the legacy quantification mode to see if that creates a fragment quant column.

But in the meantime, here are all the column names from the report.parquet file using default settings:

I'm happy to send you the output. Is there a good place that I could send it to?

[vdemichev]

Dear MSstats Team,

Wanted to ask a question about making PTM output compatible, I guess makes sense to put it here.

First, to comment about fragments: one can direct DIA-NN to output fragment quantities, scores and IDs in the main report with --export-quant. This is off by default to minimise the amount of RAM needed to store the .parquet report when loading in R - matters for large experiments.

The question I wanted to ask. What would be the most convenient tall PTM site output format (with quantities already calculated), i.e. sth similar to mstats_data$PTM? If you already happen to have a specification, could you please point me to it? I have seen examples here https://www.bioconductor.org/packages/release/bioc/vignettes/MSstatsPTM/inst/doc/MSstatsPTM_LabelFree_Workflow.html, however wanted to confirm several questions about the mstats_data$PTM representation:

1. When a precursor matches multiple proteins, is a single site on that precursor reported multipel times as distinct sites on those proteins? If no, how the single protein is chosen? If yes, is it reported for all proteins in a group or all proteins matching the precursor (i.e. very often just one protein in a group but many proteins from which precursor could originate from, based on the sequence database).
2. When a precursor matches the sequence of a protein multiple times, similarly, a single site is reported or multiple sites?
3. If a single protein site is part of multiple precursors with the same stripped sequence, is it reported only once (i.e. one row per run) or multiple times, for each precursor?
4. Likewise, if a single protein site is part of multiple precursors with different stripped sequences, is it reported only once (i.e. one row per run) or multiple times, for each stripped sequence?

Best,
Vadim

[tonywu1999]

Hi Vadim,

Apologies for the delayed response, we're just wrapping up our May Institute short course.

Anything similar to the format of mstats_data$PTM would be very helpful and convenient, especially if the PTM site location with respect to the protein sequence is in the file already. This would simplify any conversion code on our side from DIANN to MSstatsPTM format.

Below are my responses to your questions:

When a precursor matches multiple proteins, is a single site on that precursor reported multiple times as distinct sites on those proteins? If no, how the single protein is chosen? If yes, is it reported for all proteins in a group or all proteins matching the precursor (i.e. very often just one protein in a group but many proteins from which precursor could originate from, based on the sequence database).

Currently, if a precursor matches multiple proteins, MSstats removes it from the analysis. This is because it's ambiguous how much of the precursor intensity corresponds to each potential source protein. For DIANN outputs, we typically use the Protein.Names column to identify the protein, but please let me know if there’s a more appropriate column we should use.

When a precursor matches the sequence of a protein multiple times, similarly, a single site is reported or multiple sites?

Just to clarify—do you mean a case where the precursor sequence appears more than once within the same protein? For example, if the precursor is ABC and the protein sequence is ABCABC, the match occurs twice. In such cases, MSstatsPTM excludes the precursor because it’s unclear which site of the sequence is modified.

If a single protein site is part of multiple precursors with the same stripped sequence, is it reported only once (i.e. one row per run) or multiple times, for each precursor?

In this case, it is reported multiple times, one row per precursor.

Likewise, if a single protein site is part of multiple precursors with different stripped sequences, is it reported only once (i.e. one row per run) or multiple times, for each stripped sequence?

It is reported multiple times, one row per precursor.

[vdemichev]

Hi Tony,

Many thanks for the thorough explanation!

Just to clarify—do you mean a case where the precursor sequence appears more than once within the same protein?

Yes, exactly.

In this case, it is reported multiple times, one row per precursor.

OK, this is convenient, will include in the next release :) Btw, how do you deal with this afterwards, i.e. what's the algorithm for 'quantifying the site'?

Best,
Vadim

[tonywu1999]

Also, for reference DIANNtoMSstatsPTMFormat has a vignette with example data from PRIDE ID PXD053502 where ubiquitination is measured.

ptm-format.csv - Data in MSstatsPTM format

annot.csv - experimental design

report.csv - DIANN report

[vdemichev]

Super, thanks!

[tonywu1999]

If a single protein site is part of multiple precursors with the same stripped sequence, is it reported only once (i.e. one row per run) or multiple times, for each precursor?

In this case, it is reported multiple times, one row per precursor.

OK, this is convenient, will include in the next release :) Btw, how do you deal with this afterwards, i.e. what's the algorithm for 'quantifying the site'?

To clarify, I'm assuming that by "multiple precursors," you’re referring to precursors with the same stripped sequence but different charge states—for example, one with charge 2 and another with charge 3. Please correct me if I’ve misunderstood.

In this case, MSstatsPTM treats each precursor as a separate feature corresponding to the same PTM site. These features are then summarized into a single value using Tukey’s Median Polish during the MSstatsPTM summarization step.

[vdemichev]

To clarify, I'm assuming that by "multiple precursors," you’re referring to precursors with the same stripped sequence but different charge states—for example, one with charge 2 and another with charge 3. Please correct me if I’ve misunderstood.

Yes. Or also PEPT(Ph)IDES(Ph)SK or PEPT(Ph)IDESS(Ph)K both can be used to quantify the T(Ph) site.

Tukey’s Median Polish during the MSstatsPTM summarization step

Thanks!

[tonywu1999]

Yes. Or also PEPT(Ph)IDES(Ph)SK or PEPT(Ph)IDESS(Ph)K both can be used to quantify the T(Ph) site.

Actually, that one is an interesting case. In that scenario, MSstatsPTM treats them as different proteins. In the above example (treating that peptide as the beginning of the protein sequence), the first protein has T(Ph) at site 4 and S(Ph) at site 8, the second protein has T(Ph) at site 4 and S(Ph) at site 9.

To be honest, it's a bit unclear if these double modifications should be treated as different or the same proteins, and most likely depends on the user's biological question of interest. To some extent, MSstatsPTM's converters allow users to adjust if these are treated as separate or the same protein