Problems processing Waters MS/MS for GNPS

[Josh Burns]

Hi all,

I'm trying to use MZmine as part of the GNPS workflow for my Waters files.  Currently trying it on a single actinomycin standard so I'll know if it's worked properly, but run into a few problems.  Viewing just MS2 on the raw data file works, so the MS DDA acquisition shouldn't be at fault.

Files have been converted to .mzxml using GNPS' file converter, though also run into problems just using MSConvert.  I'm following the Feature Based Molecular Networking instructions, can attach the mzxml's if it'll be any use.  Mass detection set to just 1 for MS1 and MS2.

First issue is after Mass Detection on MS levels 1 and then 2, I get a "Scan #1 does not have a mass list" error, presumably because the UV wasn't mass detected.  Short of redoing my samples without UV detection, has anyone else had this problem?

Secondly, if I filter out UV (MS0) with MSConvert, then when it comes to the filter stage "Keep only peaks with MS2 (GNPS)", everything's removed.  Looking at the mass lists, all of the peak rows have "Show MS/MS" greyed out.  Guessing this is part of the problem.  Any suggestions on what I'm doing wrong?

Thanks,

Josh

[Kyo Bin Kang]

Hi Josh,

Nice to meet you as a previous Waters-user for GNPS. There are several ways to solve this problem,

If you use Mzmine for feature-based workflow, the easiest way to do it is just importing the .raw data to Mzmine directly. You can see that UV spectra would be identified as 'opposite' polarity to the real MS data, so you can easily remove them using polarity crop filter in Mzmine.

Another way to do this is... if you open the .raw data (actually it's a folder, you know), you can see sets of files named _FUNC001, _FUNC002... etc. These contain data from each 'functions' of LC-MS runs, including UV and Lockmass channel. You can identify each channel using MassLynx, and by deleting unwanted channel and converting them into mzXML, you can remove UV data from your file.

Please try these and let me know if you have further problem in converting waters data.

Kyo BIn

[Josh Burns]

Hi Kyo Bin,

Thanks for your reply.  I've imported the .raw file in, unfortunately UV spectra aren't identified as opposite - I ran in positive mode, and everything's showing up as positive, including UV.  Filtering by scan definition *TOF* doesn't work as then the MS1 and 2 scans are just overlaid, giving the "scan x has a higher retention time than the previous scan" error.  Same error after deleting the _FUN files.

I removed the UV from the GNPS convert by filtering for scan definition "Full".  This does let me mass detect 1 & 2 and build the peak list, but again when filtering for "Keep only peaks with MS2 (GNPS)", the list comes up empty with MS/MS greyed out.

Thanks,

Josh

[isabelle...@gmail.com]

Hi

I also have problems to find the right parameters to process waters data with mzmine preprocessing as advised in the documentation.

I work with a synapt G2 in survey mode. My chromatographic peak do not look so nice as the instrument switches from MS et 3 MS/MS acquisitions to acquire the 3 most intense MS/MS spectra for each scan. Did you already implemented the workflow for this instrument and what could be the recommended parameters for each mzmine process step?

I start with mass detection (MS level 1 and MS level 2) then ADAP chromatogram builder followed by chromatogram deconvolution and I have difficulties to find the most intense peak I can see when I look at the raw data.

In addition, the mass list name should be the same for both mass detection level 1 and level 2?

Thanks

Isabelle

[Sacha Pidot]

Hi all,

Not an expert MS user but have been collecting data on our recently purchased Waters Vion and have been trying to use it in GNPS. I've experienced a few errors (similar to both Josh and Isabelle) and have been able to hack a workaround (at least I think it works as I can upload the data and create a network that looks reasonable). Our machine uses the UNIFI software system, so this info only applies to that system. If you have an older Waters machine and can actually see and select your raw data files (like in Windows Explorer or similar) then half your luck - UNIFI hides the raw data files within an oracle database so they are only accessible via exporting.

For what it's worth, I think I have had some success with MSe data and GNPS. I have outlined what I have done below, but perhaps someone can comment on whether this is reasonable:

1) Collect data using MSe mode and without ion mobility. Samples collected using ion mobility cannot be exported to .raw format through UNIFI - they fail every time. The local Waters rep tells me this will be fixed in the next UNIFI update, but no time line.

2) Use "process" feature within UNIFI to peak pick and combine spectra - in Waters language this is called "componentization".

3) View data in UNIFI and turn off all filters. Select required files and export as .mzml (this will take a long time!). When I have tried to import these .mzml files directly into GNPS it has not worked and I get the "Scan #1 does not have a mass list" that Josh described above.

4) Convert to .mzxml using GNPS file converter. Remove all square brackets, spaces, etc from file names before conversion. I know this sounds like it might not make a difference, but it was the only way I could convert my files.

5) Upload to GNPS, sit back and enjoy your nice network.

As I said, I'm not an MS expert and perhaps this approach is too reductionist - does all the converting muck up your data? I haven't had time to do any sort of comparison yet.

If all else fails, you could always collect in DDA mode - if you are a UNIFI user you should be able to export directly to .raw format when data is collected in DDA mode.

Sacha

[Josh Burns]

Hi Sacha,

Thanks for that - I'm not using UNIFI so can see all the .raw files.  I thought MSe was incompatible though?  Obviously if you've got it to work then it isn't!  So if I'm just using a standard non-UNIFI machine, you'd recommend:

1. Convert .raw to .mzml with MSConvert (if so, really with zero filters?)

2. Convert the .mzml to .mzxml (why's this needed?) with GNPS converter

3. Upload etc.

Does that seem right?

Josh

[Sacha Pidot]

Hi Josh,

As, I said, I'm no expert, I've just spent a lot of time playing around and wouldn't wish that amount of time to be wasted for anyone else!

I'm afraid I can't really comment on a non-UNIFI system, as we only have a UNIFI-based machine. All I know is that we cannot export data containing IMS in .raw format using UNIFI in order to then convert to .mzxml. The only way to export MSe and IMS data from UNIFI is to process with the Waters algorithms, then export the procesed data - the only available export formats for processed data in UNIFI that will work with GNPS are .mzml or .mgf. 

I have to correct my previous post here - you can collect data using ion mobility and MSe for GNPS as long as it is processed by UNIFI and then exported as an .mzml file (I suspect the IMS data is removed when exporting in .mzml anyway). For DDA data without IMS in UNIFI, you should be able to export directly into .raw format. Processing MSe data with UNIFI apparently allows a precursor mass to be "calculated" by the software, is my understanding and how it was explained to me.

As far as the double conversion is concerned, all I know is that if we don't do it, GNPS crashes with the "Scan #1 does not have a mass list" error - I have not idea what exactly the conversion does - I agree it doesn't make sense - but for us it seems to work. I haven't played around with data collected using DDA and then exporting in .raw format, but plan to in the future.

Really sorry I can't be of more help, but maybe give the double conversion a go (as you suggested) and see what happens. 

Kind regards,

Sacha

[Pieter Dorrestein]

I suggest to contact Jimmy Yuk at Waters, he may have a solution.

P

[Kyo Bin Kang]

Hi Josh,

I forgot to mention this:

Mzmine will organize your spectra in an order of MS1 - MS2 - lock spray if you used it - UV, and each 'channel' will be sorted according to retention time.

Basically, you can evade that problem by setting Mzmine parameter 'scan numbers' when you use chromatogram builder module.

For example, see files here:

https://www.dropbox.com/sh/4paiwc6lyj7cq0z/AAA-aCK8f2Tu83eojgiwcYZha?dl=0

If you import rawdata.raw to Mzmine, you can see spectra from each function:

scan number 1 - 1022 : MS1

scan number 1023-7152: MS2 (DDA acquired from three most intense MS1 ions)

scan number 7153-7351: lock mass spray

scan number 7352-31352: UV

Basically you can build chromatogram with MS1, by setting scan number range from 1 to around 5000; it will prevent the "scan x has a higher retention time than the previous scan" message. Otherwise, if you import Data_without_UVFunc.raw, in which UV and Lockmass function files were removed, you can see that now scan numbers 7153-31352 were removed. I tested mzXML convert with this file, and it worked well; so my recommendation is just delete the UV Function files and convert the file into mzXML, using our double-click converter (https://ccms-ucsd.github.io/GNPSDocumentation/fileconversion/).

If you still have any problem, then please upload your .raw or mzmine files; I'll inspect on them.

Kyo Bin

2018년 7월 24일 화요일 오전 2시 51분 22초 UTC-7, Josh Burns 님의 말:

[Josh Burns]

Hi Kyo,

I tried deleting the UV functions, converted and ran through GNPS - still not getting an actinomycin (or any other compound) match.  Converting the .raw using MSConvert does generate some clusters, but again no compound identification.  I guess it's possible my original DDA settings were too permissive so there's more noise than signal in this file - if you could have a look it'd be great, the .raw is here https://drive.google.com/drive/folders/1xIgjw76tdyr6_X5HyNUF045dxzwTwJGE

Thanks and have a good weekend!

Josh

[Kyo Bin Kang]

Hi Josh,

I could run the GNPS by deleting the UV func and coverting to mzXML; however, as you described, I also found that there're only some singletons without any library match, even with the parameter setting of MMP 2 (https://gnps.ucsd.edu/ProteoSAFe/status.jsp?task=31654c753b614b36a3e395b68441564a). I guess this problem did not originate from the Waters format, but from the DDA data acquisition quality. See the attached PDF, then you can find that your data has too long time-interval between MS/MS scans; maybe that can be a reason why MS/MS scans were not clustered. I also found that the MS/MS scans have too low number of ions detected.

Kyo Bin

2018년 8월 3일 금요일 오전 10시 12분 30초 UTC-7, Josh Burns 님의 말:

[Neelutpal Gogoi]

Hi Kyo Bin,

Is the file conversion regarding Water's raw data has been solved or any specific protocol has been developed? I have done some LC MS/MS analysis for plant extract fractions in Waters UPLC TQD system, so now when I gone through the documentation of GNPS for analyzing the data I have observed the draw back with Waters file.

Thanking you.

[Kyo Bin Kang]

Hi Neelutpal,

As I wrote before, you can directly remove UV and rockspray data. Now this is contained in the GNPS file conversion documentation (https://ccms-ucsd.github.io/GNPSDocumentation/fileconversion/).

Another way to do it is using Symphony, which is a software Waters is selling. Related information can be found in the GNPS doc, too.

If you have any problems in converting or analyzing data, please let me know.

Kyo Bin

2019년 10월 26일 토요일 오후 2시 53분 11초 UTC+9, Neelutpal Gogoi 님의 말:

[Neelutpal Gogoi]

Thank you. If I find any problem, I'll inform you.

[Tanya Clements]

Hi all,

Has this issue been resolved? I have been struggling to convert my Waters raw files (obtained using LC-MS^e) to mzXML format using MSconvert (proteowizard).

 

I have attempted to delete certain “FUNC” files to see if this helps with the conversion and subsequent networking analysis on GNPS. However, I have had no luck.

 

Has anyone successfully used LC-MSe data and converted to mzXML format? Or do I perhaps need to use Symphony waters product for the data processing?

 

Thanks

Tanya

[Kyo Bin Kang]

Hi Tanya,

Yes, basically molecular networking accepts data acquired in DDA mode. You need to use Symphony for applying molecular networking to the MSe data.

The other option is using MS-DIAL and feature-based molecular networking workflow. You can find relevant information on the GNPS documentation pages, https://ccms-ucsd.github.io/GNPSDocumentation/featurebasedmolecularnetworking/  and https://ccms-ucsd.github.io/GNPSDocumentation/featurebasedmolecularnetworking-with-ms-dial/.

2020년 7월 3일 금요일 오후 4시 23분 25초 UTC+9, Tanya Clements 님의 말:

[Yonghui Dong]

Hi Tanya,

For Waters MSE raw file,  You need to convert channel 1 and channel 2 separately using MSConvet. Alternatively, MS-DIAL can be also used to prepare files required for GNPS. With Waters MSE data, the problem is not conversion but deconvolution. 

Waters Prognosis QI software could deconvolute MSE data. Alternatively,  MS-DIAL is also good choice. 

Based on my personal experience and mails exchange with GNPS people, GNPS does not currently support DIA data. You molecular network will be clusters of peaks and in-source fragment at the same RT, which is not meaningful. What I can think of is to remove the in-source fragment from channel 1, and slice the channel 2 data accordingly. R package RAMClust might be helpful.

Dong

--
You received this message because you are subscribed to the Google Groups "GNPS Discussion Forum and Bug Reports" group.
To unsubscribe from this group and stop receiving emails from it, send an email to molecular_networking_b...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/molecular_networking_bug_reports/309ed0cc-452e-48bd-9804-3072a75fe607o%40googlegroups.com.