mixed mode search in DIA-NN

[weiclav]

Hi there,

I would like to know what is the best way to set up a mixed search mode in DIA-NN, i.e. to use library from e.g. gas phase fractionation (GPF) experiment and to use also provided FASTA files to check for the peptides not covered in the library.

I am currently experimenting with addition of GPF data to the individual samples measurements with MBR enabled, but the improvements are quite small compared to the library-free search mode (it may be potentially related to the fact the data comes from TIMS).

Thank you in advance!

Best,
David

[vdemichev]

Hi David,

What one can do is analyse your DIA experiment and your GPF runs in a single go in library-free mode - to create a spectral library. Then reanalyse the DIA experiment with this library. Filter the output based on Q.Value, Lib.Q.Value and Lib.PG.Q.Value. The advantage of this over MBR would be that you can use the 'Unrelated runs' option during the library-creation stage only, allowing, for example, DIA-NN to use different scan window settings for different runs. But if the optimal scan window/mass accuracies are the same for DIA and GPF runs, yes, can just use MBR.

Yes, improvement in comparison to pure library-free might be small, because library-free in the latest DIA-NN is exceptionally good.

Alternatively, can create a library from your DIA data, and then just load it along with the GPF library, using --lib command twice. But the way to do it as described above is I think preferable.

Best,
Vadim

[vdemichev]

Hi David,

A note: whenever you use MBR, you always automatically get also a non-MBR report ('first-pass' files). So MBR is just something extra DIA-NN does. So if in doubt, just run with MBR and then compare the performance with/without, based on the respective reports.

1. Yes, MBR is strongly recommended for quantitative library-free searches. In this case it offers a tremedous improvement. Yes, Lib.Q.Value, Lib.PG.Q.Value, also Q.Value (if set to >1% in the GUI) and (optionally) PG.Q.Value (run-specific protein q-value, unique to DIA-NN, see the FAQ on FDR filtering).
2. I suggest to always use .tsv for any 'serious' analyses. It's better for reproducibility.
3. Yes, for very-specific library MBR usually need to be switched off. For GFP - that's almost certain. For DDA-based project-specific libs I have seen substantial improvement with MBR in some cases.

Best,
Vadim

[weiclav]

Thank you very much for the update!

Take care,
David

[weiclav]

Hi Vadim,

sorry to bug you with this one once more. I am still uncertain with the main report filtration when the DIA sample runs and GPF runs libraries are used and MBR is not checked or checked but first pass report is used, i.e. 3rd scenario mentioned here.

Should one use Global.x or Lib.x q values instead in a situation where you have project specific libraries so big part of the library is being reported? You have mentioned in filtering on these three levels before in your first response here: Q.Value, Lib.Q.Value and Lib.PG.Q.Value; is this correct?

Thank you again!

Best,
David

[vdemichev]

Hi David,

If you don't use MBR, use q-values marked as 'Global' for global filtering. If you use MBR, use q-values marked as 'Lib' for global filtering.

Best,
Vadim

[weiclav]

Hi Vadim,

thanks for the confirmation!

Best regards,
David