DIA-NN Phospho search issues

[JHKC12]

Hi there,

I'm trying to run 4x phospho enriched and 4x global samples in DIA-NN (1.8.2 beta 39) but i am not sure if DIA-NN is bugged or if its still searching and taking a long time.

From what i have read in the forums, the workflow for phospho dia in DIA-NN is briefly as follows:

• input phospho and global raw files

• add fasta (in this case, mouse)

• check both boxes under "precursor ion generation" ('FASTA digest for lib free' and 'deep learning based spectra')

• missed cleavages 1

• max no. variable mods 3

• phospho checked

• precursor charge range 2-4

• library generation as "IDs, RT and IM profiling"

• MBR checked

• other settings as default

• when this has finished, perform a second DIA-NN search with the spectral library generated with similar settings

However, i am running the first search and the current log for the search is below:
[0:00] Loading FASTA C:\Users\mproteomics\Desktop\Analysis Data\Sequences Databases\uniprot-proteome_UP000000589_MOUSE__55,086 (Dec 2023).fasta
[0:27] Processing FASTA
[4:13] Assembling elution groups
[7:39] 50147387 precursors generated
[7:39] Gene names missing for some isoforms
[7:39] Library contains 54858 proteins, and 22143 genes
[8:56] Encoding peptides for spectra and RTs prediction
[13:04] Predicting spectra and IMs

I have been running this for about 2 days now and i am now sure if this is a DIA-NN bug or if i am doing something wrong

is there a workflow that explains how to perform phospho samples?

thanks in advanced

[vdemichev]

What's the amount of RAM in the system? Not enough RAM could be the only reason why it's slow in this case.

[JHKC12]

Hi Vadim,

we have about 120gb RAM. I am in the process of getting the PC upgraded to more processing power.

But is the workflow above the correct way to go with phospho DIA?

[vdemichev]

We describe exact recommendations for phospho here now:
https://github.com/vdemichev/DiaNN?tab=readme-ov-file#ptms-and-peptidoforms

[JHKC12]

We describe exact recommendations for phospho here now: https://github.com/vdemichev/DiaNN?tab=readme-ov-file#ptms-and-peptidoforms

Hi Vadim,

we upgraded our PC with higher processing power and also used the latest version of DIA-NN with the settings used as per the recommendations provided (using 'ultrafast' as well). However, the search seems to always be stuck at the step "[16:54] Predicting spectra and IMs".

Is it common for this step of the workflow to take long in the generation of the spectral library with 3 variable modifcations?

[vdemichev]

On a Ryzen 7950X (16 cores) or 10980XE (18 cores) it takes about 3 minutes per million precursors. That is, a rough time to generate a 50-million precursor library on such a PC would be ~150 minutes - with DIA-NN 1.9. So no, it should not take much longer. Can you please share the full log generated so far using the 'Save log' button?

[JHKC12]

On a Ryzen 7950X (16 cores) or 10980XE (18 cores) it takes about 3 minutes per million precursors. That is, a rough time to generate a 50-million precursor library on such a PC would be ~150 minutes - with DIA-NN 1.9. So no, it should not take much longer. Can you please share the full log generated so far using the 'Save log' button?

heres the log saved so far:

---

Skyline not found
MSFileReader found: MSFileReader Core 31

diann.exe --lib "" --threads 32 --verbose 1 --out "C:\Users\mproteomics\Desktop\Analysis Data\Sequences Databases\SpecLib_FREE\Mouse_Phos.tsv" --qvalue 0.01 --matrices --out-lib "C:\Users\mproteomics\Desktop\Analysis Data\Sequences Databases\SpecLib_FREE\Mouse_Phos-lib.tsv" --gen-spec-lib --predictor --fasta "C:\Users\mproteomics\Desktop\Analysis Data\Sequences Databases\uniprot-proteome_UP000000589_MOUSE__55,086 (Dec 2023).fasta" --fasta-search --min-fr-mz 200 --max-fr-mz 1800 --met-excision --min-pep-len 7 --max-pep-len 30 --min-pr-mz 300 --max-pr-mz 1800 --min-pr-charge 2 --max-pr-charge 4 --cut K*,R* --missed-cleavages 1 --unimod4 --var-mods 3 --var-mod UniMod:21,79.966331,STY --peptidoforms --reanalyse --relaxed-prot-inf --rt-profiling
DIA-NN 1.9 (Data-Independent Acquisition by Neural Networks)
Compiled on Jun 8 2024 20:00:31
Current date and time: Mon Jun 24 13:14:29 2024
CPU: GenuineIntel Intel(R) Xeon(R) Gold 6130 CPU @ 2.10GHz
SIMD instructions: AVX AVX2 AVX512CD AVX512F FMA SSE4.1 SSE4.2
Logical CPU cores: 64
Thread number set to 32
Output will be filtered at 0.01 FDR
Precursor/protein x samples expression level matrices will be saved along with the main report
A spectral library will be generated
Deep learning will be used to generate a new in silico spectral library from peptides provided
Library-free search enabled
Min fragment m/z set to 200
Max fragment m/z set to 1800
N-terminal methionine excision enabled
Min peptide length set to 7
Max peptide length set to 30
Min precursor m/z set to 300
Max precursor m/z set to 1800
Min precursor charge set to 2
Max precursor charge set to 4
In silico digest will involve cuts at K*,R*
Maximum number of missed cleavages set to 1
Cysteine carbamidomethylation enabled as a fixed modification
Maximum number of variable modifications set to 3
Modification UniMod:21 with mass delta 79.9663 at STY will be considered as variable
Peptidoform scoring enabled
A spectral library will be created from the DIA runs and used to reanalyse them; .quant files will only be saved to disk during the first step
Heuristic protein grouping will be used, to reduce the number of protein groups obtained; this mode is recommended for benchmarking protein ID numbers, GO/pathway and system-scale analyses
The spectral library (if generated) will retain the original spectra but will include empirically-aligned RTs
Exclusion of fragments shared between heavy and light peptides from quantification is not supported in FASTA digest mode - disabled; to enable, generate an in silico predicted spectral library and analyse with this library
The following variable modifications will be scored: UniMod:21
WARNING: MBR turned off, two or more raw files are required

0 files will be processed
[0:00] Loading FASTA C:\Users\mproteomics\Desktop\Analysis Data\Sequences Databases\uniprot-proteome_UP000000589_MOUSE__55,086 (Dec 2023).fasta
[0:34] Processing FASTA
[5:39] Assembling elution groups
[10:38] 50147387 precursors generated
[10:38] Gene names missing for some isoforms
[10:38] Library contains 54858 proteins, and 22143 genes
[11:22] Encoding peptides for spectra and RTs prediction
[16:54] Predicting spectra and IMs

---

[vdemichev]

Thanks! Still, could it be that the RAM is all full? What is the physical RAM occupied amount shown by the Task Manager? This would be an explanation for it taking a very long time.

[JHKC12]

it says its only using 5-10% of memory

[vdemichev]

50 million database will be tens of gigabytes, strange. I would try to restart DIA-NN and if takes long see what's the reported RAM consumption.

[JHKC12]

i restarted DIANN and changed a few settings:

• precursor charge range from 2-4 to 2-3
• max number of variable mods from 3 to 2
• increased thread count to 50

there has been further progress and is now saying:

0 files will be processed
[0:00] Loading FASTA C:\Users\mproteomics\Desktop\Analysis Data\Sequences Databases\uniprot-proteome_UP000000589_MOUSE__55,086 (Dec 2023).fasta
[0:21] Processing FASTA
[2:19] Assembling elution groups
[3:32] 20805637 precursors generated
[3:32] Gene names missing for some isoforms
[3:32] Library contains 54858 proteins, and 22143 genes
[3:43] Encoding peptides for spectra and RTs prediction
[4:50] Predicting spectra and IMs
[926:24] Predicting RTs

this has been running for about 15 hours now so i am not sure if this is still taking too long. the memory usage is now at around 30%

[vdemichev]

This is indeed quite strange. The CPU load in Task Manager does correspond to what you'd expect, based on the number of threads set, and no other high-CPU tasks are run on the machine at the same time?

[JHKC12]

This is indeed quite strange. The CPU load in Task Manager does correspond to what you'd expect, based on the number of threads set, and no other high-CPU tasks are run on the machine at the same time?

Not that i was able to see. we managed to get through the data but it looks to take a lot longer than what others have experienced