No protein identification despite good chromatograms/signal

[joachims1956]

Dear DIA-NN team,

Currently we are running DIA-NN with TimsTOF SCP acquired samples. We get very good signal and clean spectra. Also the calibration status of the instrument is good (below 10ppm). Nevertheless, we do not get any protein identifications. Are there any additional DIA-NN commands we could try for troubleshooting. So far, we:

• tested different libraries
• relaxed protein inference settings
• applied different FRA (1 - 10)
• tested different mass calibration accuracy values (10 - 1000)
• broader IM and RT ranges
• with/without fragment selection (--no-fr-selection)
• with/without interference subtraction from fragment ion chromatograms (--no-ifs-removal)

Very much thanks for consideration. Best

[vdemichev]

Hi Joachim,

What is the nature of the sample? Can you please share the log obtained at log level = 3 when using some public spectra library suitable for this kind of sample with (important) deep learning activated in DIA-NN?

Best, vadim

[joachims1956]

Dear Vadim,

Thank you for your reply. The samples of the logfile are fresh stained single cell samples of mouse tumours. But we get the same results/no identification also with fresh unstained murine single cell samples. We also tested different mouse spectral libraries including the DIA-NN in-silico FASTA digested one.

Please find added the logfile.

To us, it looks that the samples are quite empty and that DIA-NN could get only low-confidence identifications. We would be happy to learn whether we still could get some output of these files by adjusting DIA-NN settings. Or whether we could pinpoint the reason for these "no identifications". In Hela single cell samples we are getting several thousands protein identifications per sample.

Thank you very much. Please let me know if you need additional information of in case of questions. Best report_FITC_stainded_mouse_single_cells_in_silico_sepclib_two_samples_17042024.log.txt

[vdemichev]

Here, just no proteins pass the 1% FDR threshold. Considering the very low numbers of detectable precursors, looks like just barely anything is detectable. Common contaminants can also give a 'good-looking' XIC. Mass calibration is indeed fine here. One could increase the number of IDs by using a spectral library (any kind of mouse lib), but this is unlikely to solve the problem. By the looks of it, either the target proteins are not there, or there's very strong background signal.

Best, Vadim