Skip to main content
Topic: Consequences of using Centwave for profile mode data (Read 3279 times) previous topic - next topic

Consequences of using Centwave for profile mode data

Dear All,

I am trying to reproduce/reprocess a recently published data (data was open to public) and I found out that the author used centwave algorithm to process profile mode data (version of R used 3.2, she used a wrap around function for xcms peak picking so I can't tell which version of xcms she used). I would like to know if there will be any negative effect on the final data table using her method. As far as I know, xcms online will not accept using centwave for profile mode data?

Many thanks for your help

Flo

Re: Consequences of using Centwave for profile mode data

Reply #1
Gusses of what will happen:
* She uses a wide mz window in peak picking to get the whole peak inside. In effect treating it like it has much less resolution
* She uses a normal mz window (ppm) and each peak is split in many pieces. Each real peak is represented by many features in her feature table. Probably there is many more features than you would expect.

It is hard to imagine you could do this without noticing. Perhaps they put the raw data up but forgot to say that they centroided?
Blog: stanstrup.github.io

Re: Consequences of using Centwave for profile mode data

Reply #2
Dear Jan,

Many thanks for your reply. That was what I guessed before...so, I emailed the author specifically asking her about her profile data and if she would at least give me her peak picking settings.

According to her, she used 2ppm (I honestly surprised) , mzdiff = 0.001 and she did not centroid her data.....and reported 800 analytes in her paper.

I tried to repeat her settings in xcms, it has taken 2 days now and it's still running (when I used some sensible settings, 10 ppm and  convert to centroid data, it only took a few hours).

What I am really worried about is the final data quality as I need to integrate these results with transcriptomics data. I would like to be accurate if I need to criticise her result in my MSc project.

Many thanks.

Flo




Re: Consequences of using Centwave for profile mode data

Reply #3
I would ask for here complete settings at this point and see if you can replicate at all. the prefilter setting in particular might explain how she was able to run the analysis at all. The grouping step might also have put humpty dumpty somewhat back together again but you should see a huge npeaks number in the peak table. a positive "mzdiff" I guess also might join things up to look sensible on the surface.
Blog: stanstrup.github.io