DIA-NN 1.9, different results for theoriticaly same processing
NicoDrou opened this issue · comments
Dear Vadim,
Based on your advices, I have generated the predicted spectra in a separate pipeline before i process the data with this library and reannotate it with the same FASTA file I used.
The 2 predicted spectral library have the same size, so i guess we can assume they are identical (?).
And to compare, I did the same with the same parameters but in one step, and the results are very different.
In the PG matrix of the 2 steps I detect 512 PG, and only 420 in the one step.
I can't explain why. Is it something that is expected?
I enclose here the log file of the different step
Step 1_Generation library.txt
Step 2_Data processing.txt
Step3_ALL-in-ONE.txt
Hi Nico,
The 'all in one' method is not supposed to produce the same results, but rather is expected to have 'comparable' performance. Which is the case here. There are certain differences how this 'on the fly' search is handled at the moment - I will likely remove those in the future, but for now it's different. Further, in this case there's also a difference due to M(ox) being declared as a var mod in one case (all in one) but not the other - this leads to PTM localisation enabled for M(ox) only in all in one case, which changes things too.
In general, please use the two-step procedure in all cases, I have amended the docs for DIA-NN 1.9 to make it the guideline.
Best,
Vadim
Hi Vadim,
thank you for your super fast answer ! :)
well indeed the M(ox) was declared in the all in one, but the library was created with the M(ox) declared at the first step. Does it means Precursor Generation section also needs to match ? i thought this section was only used for the spec-lib generation and then all the MSMS spectra were searched against the predicted spec-lib.
Does it means Precursor Generation section also needs to match
Just need the respective M(ox) checkbox checked. This changed in DIA-NN 1.9 in comparison to 1.8.1. I guess I need to rename this section in the GUI :)
ok ok. thank you very much ! I will reprocess the data with the respective M(ox) checked,
However, I just checked in the 2 steps processing and I have many precursors containing the UniMod:35.