I am analysing LC-MS data. I have combined all NEG and POS mode peak intensities in a single file retaining all metabolites with higher peak intensities (appearing in either modes). Now, when I go for analysis in MetaboAnalyst, which ones should I select among:
1. Sampe normalization
2. Data transformation
3. Data Scaling
Thank you for guidance!
But why do you want to combines NEG and POS files, if the voltages, lenses adjustment, dry gas temperature and other settings are different?
I would treat as separate samples (files) and compare them.
The mz values for the same molecule in NEG and POS modes are different at MS1. IF (a big if) you assume that everything is protonated in POS mode, you can, in principle reduce the peak list back to the zero charge state. This is easier in MALDI (singly-charged ions) than ESI (multiply-charged ions), but it can be done, at least in principle. IF (again, big if) you assume that everything is proton deficient in NEG mode, then you can create a zero charge spectrum of the NEG mode. This will create a master list of MS1 molecular ions (some will be seen in POS mode and others in NEG mode, with most seen in both modes). At the MS2-level this no longer works since the fragmentation patterns are completely different for POS ions and NEG ions.
If you were to try to concatenate the data files into a single LC-MS/MS data set. it will require some knowledge of programming and MSXML data structures, but it is doable. If you were to interleave the POS and NEG spectra by elution times, then you will destroy the continuity of peaks, making most appear as noise. Worse, you would generate lots of nearly isobaric overlaps that would be impossible for the software to properly resolve. If you were to append one mode on the other as contiguous datafiles, then the elution times would overlap and confuse the software. You could add a fixed time, beyond the ending elution time of the first mode to the elution times of the second mode to fool the software that the second set of peaks were just an extension of the same run. This would be the only way to keep the mz overlaps from occurring. Note, the background noise levels are likely to be different between the two runs, so you would need to do some baselining separately before the merge.
Now the issue is what statistical benefit would this offer to the analysis? I think the better result is to run your PCA analysis on both sets independently, then look for those principle features that replicate between the two runs (i.e., semi-independent confirmation).
Neg and Pos mode data are created from two different methods. Do not combine different methods. Each one should be shown separately for context, they are not additive.
yes you can do that if you are doing untargted metabolomics or lipidomics. But, first you should have a software for normalisation, allighnment, and peak picking. The software may be provided with your vendor if you use Agilent or waters or Thermo instrument.
Thank you very much @ Ahmed; William; Afroz; and Thais for your answers.
I reckon the combination of NEG and POS data will ruin the downstream statistical analyses for many research questions like biomarker discovery. I won't do that. Pareto scaling, quantile normalisation and log transformation are the most popular, but different data may require different data prep before stats analyses.!
Kindly visit...
Article A relative quantitative positive/negative ion switching meth...
- Badges
- Science topic
- Similar topics
- Biological Science
- Anatomy
- Tissue