Open WaiChingSin opened 10 months ago
Hi Johnson,
Predicting 32 million spectra will be very slow with older DIA-NN versions, please try 1.8.2 beta 27 where this issue is fixed. Also, the next major update with much better phospho will be out soon.
Best, Vadim
Hi Johnson,
Predicting 32 million spectra will be very slow with older DIA-NN versions, please try 1.8.2 beta 27 where this issue is fixed. Also, the next major update with much better phospho will be out soon.
Best, Vadim
How could I use DIA-NN for searching N-phosphorylation(H, R, K)?
Dear Vadim,
I am interested in performing quantitative phosphoproteomic using library-free mode. The library-free mode is fast when doing unmodified peptide search but I am concerned about the search time when doing phosphopeptide search. We may also attempt to use custom fasta in our study, so it would be difficult for us to adjust parameters if the library building step take too long.
I would like to ask more about the library generation:
Here is the log of my library preparation: 0 files will be processed [0:00] Loading FASTA D:\Johnson\Databases\Gmax_508_Wm82.a4.v1.protein.fasta [0:18] Processing FASTA [2:07] Assembling elution groups [3:41] 33269738 precursors generated [3:41] Gene names missing for some isoforms [3:41] Library contains 85865 proteins, and 85865 genes [3:46] Encoding peptides for spectra and RTs prediction [5:08] Predicting spectra and IMs
I am not sure if it bugged at the "Predicting spectra and IMs" step or it simply require days(or weeks?) for the library generation.
These are the parameters for the library generation:
Thank you, Johnson