High median RT prediction accuracy values

[joachims1956]

Hello,

Thanks for developing DIA-NN. We have a questions regarding the RT prediction accuracy. In some samples we observed much higher "median RT prediction accuracy" compared to other samples of the same DIA-NN run. All samples were acquired under same LC-MS condition and DIA-NN settings. What does a higher median RT prediction accuracy mean and what can we do to avoid it? I add the diagram from the DIA-NN analysis report.

Thanks very much in advance

[vdemichev]

Means something is very wrong with those samples. I would think also almost no precursors detected in those? If yes, then the high RT prediction error is expected (i.e. handful precursors just don't allow for good RT alignment), if not - that would be very strange.

[joachims1956]

Thanks for your reply. Yes we do not see any precursors (based on DIA-NN report) I attached the screenshot below. Although DIA-NN reported MS1 and MS2 signal for those samples as well. Overall, are those DIA-NN report outputs based on very wrong DIA-NN analysis settings or some major error in lab sample processing? Thanks a lot

[vdemichev]

How do MS1 and MS2 signals reported by DIA-NN look like? The most likely explanation is either

• there's nothing in those samples or
• mass calibration is way off

To rule out the latter, can try 1.8.2 beta 11 https://osf.io/t2ymc/?view_only=7462fffb20e648fc83afc75d8c67e9f8 with --mass-acc-cal 100. Is this timsTOF? Then checking how mobilograms look like with AlphaTims definitely makes sense.

[joachims1956]

Thanks again for your reply. MS1 looks like this:

MS2 like this:

Thanks, I'll try and update.

Yes, it is timsTOF

[vdemichev]

OK, considering very low MS2, this already indicates something is wrong. Or it's supposed to be low injection amount?

[joachims1956]

Injection was, same for all samples, 20ng tryptic digest.