DDA warning / only 2 cycles detected for an RT-segmented variable-window DIA method

[Hayoung-hiro]

Hi, vadim

I ran DIA-NN on a method that applies different m/z ranges depending on retention time (following the RT-dependent DIA approach in "Increasing Proteome Depth While Maintaining Quantitative Precision in Short-Gradient Data-Independent Acquisition Proteomics", J. Proteome Res. 2023, https://doi.org/10.1021/acs.jproteome.3c00078), on an Orbitrap Exploris 480.

In the MS2-acquiring regions the cycle structure is uniform: 1 MS1 + 20 MS2 (variable width). To save scan time, the method does not acquire MS2 in void regions (RT ranges without precursors). MS1-only stretches in void regions.
As a result the overall MS1:MS2 ratio is low and non-uniform across the run (~6.5 MS2 per MS1, rather than a constant 20 per cycle), and I suspect this is what confuses DIA-NN's cycle detection.

Full log on loading the run:

[3:15] File #1/6
[3:15] Loading run C:\Users\Odyssey\Desktop\Variable_DIA_raw\Exp_raw\Batch4_HUPO\20260608_SampleA_60min_dppp7_adap_dense_01.raw
[3:19] Loaded 6730288 MS1 and 55910017 MS2 peaks
WARNING: check if this run is DDA - in this case use --dda to analyse it
[3:19] Detected MS/MS range: 99.9981 - 1999.95
[3:20] Run loaded: 7310 MS1 scans and 47284 MS2 scans in 2 cycles
[3:20] Pre-processing...
[3:25] 7310 MS1 and 47284 MS2 scans in 2 (inferred) and 2 (encoded) cycles, 7841156 precursors in range
[3:27] Calibrating with mass accuracies 30 (MS1), 20 (MS2)
[3:27] Processing batch #1 out of 200
[3:27] Precursor search
[3:35] Optimising weights

DIA-NN groups all 47,284 MS2 scans into only 2 cycles and raises the DDA warning. Changing the scan window radius (0 vs 20) made no difference.

Is this expected when the number of MS2 scans per MS1 is not constant across the run (variable windows + no MS2 in void regions), and is there a recommended way to have DIA-NN handle such methods? I'm happy to share a representative file or method details if useful.

DIA-NN 2.3.2, Thermo .raw. Thanks!

[vdemichev]

Hi,

Yes, it will not correctly detect cycles in such a setting. That said, do the outputs look good?
Btw, if you keep the covered m/z range always same (just add large windows at each end), this issue will not manifest.

MS1-only stretches in void regions

I don't understand this. If the instrument believes there's nothing to identify (and the instrument is most likely wrong), then how does it hurt to acquire MS2 anyway?

Best,
Vadim

[Hayoung-hiro]

Hi Vadim,
Thanks for the quick reply and for confirming that cycle detection won't work correctly in this setting.
On the outputs — yes, they do look good. Across these runs, protein identifications and quantification have held up well even with this acquisition setup, which is reassuring. For context, I'm developing a tool (AIDIA) that takes DIA-NN output and proposes RT-dependent m/z window optimizations, building on the short-gradient RT-dependent DIA idea from the JPR 2023 paper. On DIA-NN 1.9 the relevant per-cycle information wasn't exposed in the output, so I hadn't noticed the cycle issue; on 2.3.2 I saw the MS1/MS2 cycle miscount (2 cycles) and the DDA warning, and started digging into it.
Sorry about the unclear phrasing on "MS1-only stretches in void regions." What I meant is the void-volume region at the very beginning of the LC run, where nothing is eluting yet. I deliberately did not acquire MS2 in that stretch, reasoning that it would only inflate the file size without adding useful identifications and might affect downstream results. So that region contains MS1 scans only — and that turned out to be the main cause of the problem. The missing MS2 in those early scans is what broke the otherwise uniform 1 MS1 + 20 MS2 cycle structure and made the MS1:MS2 ratio non-uniform across the run.
After removing that MS1-only region and reordering/renumbering the scans as described by aquachdb (#1514), the cycle structure was detected correctly and the DDA warning went away.

I'll re-run the affected files with the fix and report back on whether protein ID and quantification change relative to the original outputs.
Thanks again for taking a look!
Best,