Why not 5% FDR?

[tannishthadas]

Even though I set the parameter as 5% precursor FDR, this happened :

[0:28] Processing...
[50:56] RT window set to 6.97918
[50:56] Peak width: 8.252
[50:56] Scan window radius set to 18
[50:57] Recommended MS1 mass accuracy setting: 1.63373 ppm
[65:41] Removing low confidence identifications
[65:41] Searching PTM decoys
[65:41] Removing interfering precursors
[65:42] Training neural networks: 4149 targets, 2572 decoys
[65:44] Number of IDs at 0.01 FDR: 1946
[65:44] Calculating protein q-values
[65:45] Number of genes identified at 1% FDR: 813 (precursor-level), 750 (protein-level) (inference performed using proteotypic peptides only)
[65:45] Quantification

It went ahead and considered 1% FDR only.
Is there a way of applying 5% FDR?? I need to get more proteins for my purpose which is why I am relaxing the criteria.

[vdemichev]

The log is not affected by the FDR setting. But the main report will have more precursors and might have more proteins (although not reliably detected).

[JunYao-mynamesofewletters]

Hi Vadim,

I tried different FDR (0.1, 1 and 5) and founf that only report.tsv changed while others like report.pg_matrix.tsv did not change. Will this be modified in further update?

Thank you in advance for your help.
Jun

[vdemichev]

Hi Jun,

What you observe is exactly as it is intended.

I would suggest not to change any FDR settings in DIA-NN GUI, but rather just apply the filters you want when processing the main report in R or Python.

Best,
Vadim