search

vdemichev / DiaNN

DIA-NN - a universal automated software suite for DIA proteomics data analysis.
Other
281 stars 53 forks source link

How is a precursor confidently identified? #1113

Open hxxhust163 opened 3 months ago

hxxhust163 commented 3 months ago

Hi Vadim

Hope this finds you well. Thanks for your work in DiaNN. I get a library for DDA data. Based on the intensity, I get two libraries, one is the top 12 ions and another is top 15. I run diann using the two libraries separately with same parameters. By comparing the results I found that 45 precursors only found by top 12 library and 47 precursors only found by top 15 library, while there were 2719 precursors overlap.

I have two questions: Q1. Though 45 is smaller than 2719, I wonder why this happens. It means the 45 and the 47 precursor are correct, but they will be lost if using different library.

I guess it is probably due to the FDR. The 45 precursors did not pass the FDR threshold when using the top 15 library. Q2. I wonder how the FDR calculated or how the decoy implemented.

Btw, I use diann 1.8.1 version in linux.

Best wishes Xiaoxiang

vdemichev commented 3 months ago

Hi Xiaoxiang,

45 or 47 is comparable to a random fluctation in this case, i.e. makes no difference. Note that the software never draws a clear line between 'identified' and 'not identified'. Everything 'not identified' just has q-values > 0.01.

Q2. Described in the Nature Methods DIA-NN paper.

Best, Vadim

hxxhust163 commented 3 months ago

Hi Vadim

Thanks for your reply. So, how can I get the precursors which correspond to q-value >0.01 as you mentioned? Is the param --qvalue 1 in DiaNN able to save all the precursors in the main out?

I tried --qvalue 1 but only found precursors with 0.05> q-value> 0.01 in the main out. I wonder whether the precursors whose q-value ranging from 0.05 to 1 could be saved in the main out.

Thanks very much!

Best wishes Xiaoxiang

vdemichev commented 3 months ago

Hi Xiaoxiang,

This is strongly not recommended, better to just choose a single library and use it at 0.01 q-value threshold.

Best, Vadim

hxxhust163 commented 3 months ago

Hi Vadim

Thanks! I have another question. I wonder how to have the decoys saved in the main out?

vdemichev commented 3 months ago

It's currently impossible to save decoys, however this option might be added in the future.

noahdephoure commented 3 months ago

I've always liked to keep my decoys (peptide and protein) in the dataset, until presentation. That allows me apply any subsequent filtering to them and be able to report the that fdr.