Effect of No Shared Spectra

[SamuelLozanoJuarez]

Hi,
I am using DIA-NN to test a library of modified peptides that I have created with some E. coli samples. I have read that it is recommended to leave the ‘No shared spectra’ option checked in all runs. However, I have tried running the DIA-NN search with and without ‘No shared spectra’ checked and the number of distinct peptidoforms identified is higher when this option is unchecked.

Specifically, the library I have created has ~50000 distinct peptidoforms, if I leave the ‘No shared spectra’ option checked DIA-NN identifies about 18000 peptidoforms from the library, while if I uncheck the option DIA-NN finds 38000 peptidoforms (these values are obtained from the Modified.Sequence field of the report.pr_matrix). In my case it seems to give better results without checking the ‘No shared spectra’ option.

Is it possible that I am misinterpreting the data? How does the ‘No shared spectra’ option affect?

Thank you very much.
Samuel

[vdemichev]

Hi Samuel,

Yes, number of IDs is higher with unchecked, this option is reducing IDs by primarily removing false IDs. Some peptidoform-related benchmarks here: https://github.com/vdemichev/DiaNN/blob/master/Presentations/HUPO%202024%20poster.pdf.
So the option is basically essential, there are no known scenarios when it's better to have it off.

Best,
Vadim

[SamuelLozanoJuarez]

Thank you for the quick response. Just to be sure, if a search is performed without ticking the ‘No shared spectra’ option, can peptidoforms that have been identified and pass the FDR be considered as correct? Or is there no guarantee that they are correct.

Thank you very much.
Best,
Samuel

[vdemichev]

can peptidoforms that have been identified and pass the FDR be considered as correct?

No.