Skip to main content
Topic: XCMS : Processing and the number of features picked (Read 5533 times) previous topic - next topic

XCMS : Processing and the number of features picked

Hi,

Do we get the same features, so the same diffreport, by processing 2 groups downloaded in 2 different datasets in one case, and by processing the same 2 groups downloaded in one dataset against blank samples in the other dataset ?

The same processing parameters were set in both methods.

Thank you for your answer.

Re: XCMS : Processing and the number of features picked

Reply #1
I'm not sure I understand the question correctly.

If comparison #1 is  A vs B, and comparison #2 is (A u B) vs Blank, then the result will be very different.

What are you trying to achieve ?

Re: XCMS : Processing and the number of features picked

Reply #2
Thank you for your rapi answer.

I do agree with you that the statistics will be very very different.

What I was just asking about is whether the tow matrix generated with both processing #1 is A vs B  and  #2 is (A u B) vs Blank  would be different in terms of number of features, retention time correction ...etc ?

Actually, I was given such data to analyse. I am not intending to perform comparaison on XCMS Online and to use its stat results of course. It's irrelevant since we get different results. I am just wondering if it's relevant to use the matrix generated by (A u B) vs Blank and do statistics by comparing the data A vs B by importing it on other data analysis softwares which with paying attention to exclud Blank data of course ?


Another question :

Since it seems to be a trick for a multi classe comparison not implemented on the actual XCMS release.

e.g : A, B, C, D, E, F : classes

So, could we process  (A+B+C) vs (D+E+F), then use the generated matrix for data analysis on other softwares (multivariate, pairewise analysis etc ...) e. g : A vs B , A vs F , B vs C, ... etc ?

Hope I was clear this time.

Thank you very much.

Re: XCMS : Processing and the number of features picked

Reply #3
Regarding #1 and #2 the result will be similar in terms of detected features - with the difference that XCMS will also detect the features in the blanks - and there are a lot features from contaminants, plasticisers etc in blanks.

Yes you can do statistics on subgroups, XCMS just calculates one test for you, but the diffreport contains all aligned features, not just the dysregulated ones.

One thing to keep in mind here is the minfrac parameter (alignment tab).
This parameter determines in how many samples in one group (expressed as a fraction, e.g. 0.5 = 50%) a feature has to be detected in order to consider make it valid feature.
For example if minfrac=0.5, nsamples=10 in group A and a feature is detected in only 1-4 samples it will be considered non-reproducible and will be discarded.

So you might want to change (lower) this parameter in this case (A+B+C) vs (D+E+F) according to the number of samples in the groups A .. F.

Multi-class comparison will also be implemented in XCMS Online, we are focusing on support for meta-analysis at the moment.

Re: XCMS : Processing and the number of features picked

Reply #4
Thank you very much for valuable answer.