The larger the isotope number, the larger the m/z error

[Annatkachev]

Hello,

It seems that the algorithm by which CAMERA calculates m/z between isotopes systematically leads to erroneous isotopes of higher order.

As I understand, m/z differences are calculated between isotopes [M+k] and [ M+k-1]. For higher order isotopes, it seems that there is a m/z error that is accumulated, and the higher order isotopes are in fact no longer in an appropriate m/z range to the respective [M] isotope.

To illustrate the issue, I attach differences between adjacent isotopes. The ppm error is mostly less than 10 ppm:

However, if I calculate the difference between the higher order isotopes and the [M] isotope, there are more and more peaks with large m/z differences to the [M] isotope. By the isotope [M+5], most of the detected isotopes, have, in fact, a ppm to large to be true isotopes.

While I understand that m/z values have to be calculated between the [M+k] and [ M+k-1] isotopes to define possible isotope groups for computational reasons, I think a post-processing step should be added to filter from the pre-defined groups which peaks, in fact, have appropriate m/z difference with the respective [M] isotope.

I used XCMS online to process the data, which was acquired in positive mode. The expected mass tolerance we observe is 5-7 ppm.

I attach a colab notebook with the data and analysis. I have only considered isotopes with +1 charge.
https://colab.research.google.com/drive/1OciFt1M3lKrorlX7gnkwH4skJRwnHvFV?usp=sharing

Best regards.

[sneumann]

Hi, thanks for the precise analysis! The matrix with ranges for the isotopes is defined in

CAMERA/R/fct_findIsotopes.R

Line 2 in 9f4552c

calcIsotopeMatrix <- function(maxiso=4){

For some testing, you can supply such a matrix to the function

CAMERA/R/xsAnnotate.R

Line 539 in 9f4552c

setMethod("findIsotopes", "xsAnnotate",

I also don't have the confidence intervals for m/z of the entire isotope cluster.
Maybe an implementation similar to https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3879626/ could help ?
Yours, Steffen