Q: Feature quality and isoerror

[jameswallacew8]

Hello.

I think I am correct in saying that isoerror and isoerror2 in the features output file is the mass difference in ppm between the monoisotopic peak and its 1st or 2nd isotope. Are small values close to zero desirable here as isotopes should be fairly close to the monoisotopic peak across scans? I'm not sure how the isotope mass error map fits in here. Is it used to calculate some average mass shift over all features and filter out isotopes that exceed this?

Number of scans and number of Isotopes seem to be the best indicators for feature significance / quality. Any suggestions on how to use the isoerror as well?

Many thanks,
James

[markmipt]

Hi James,

Yes, you are right. "isoerror" is the difference between first isotope and monoisotopic peaks (mz_for_first_isotope - 1.00335 - mz_for_monoisotope, in ppm). Similarly, "isoerror2" is mz_for_second_isotope - 2 * 1.00335 - mz_for_monoisotope.

In an ideal case, these values should be close to zero. However, there is exists an average mass shift because we cannot resolve isotopic fine structure due to limited resolution. Biosaur estimates this average mass shift, as well as standard deviation of mass errors and filter all isotope candidates which do not follow the condition "abs(isoerror - av_mass_shift) <= 5 * std_of_iso_mass_error".

I'm not sure that there are exist a really good indicators of feature quality. Yeah, the number of scans, number of isotopes, intensity and these isoerrors correlate with some quality. But it is also depends on what is the "quality" features for you. In my case these are the features which lead to increase in number of protein identifications when I do it using only MS1 spectra. And in my experience I see that even features with "bad" mentioned values are important, like features detected only in a single MS1 scan, having only one isotope, etc. For the quantitation, there is a lot of ambiguity in any way. However, in most cases, it seems that the best solution is to detect more features and chose the one with highest intensity.

As for the application of "isoerrors" - I've added them mostly for detection of sulfur (in C and M amino acids). There should be a higher average mass shift for second isotope in case of sulfur presense. But there is no "ready to use solution" at the moment.

Regards,
Mark

[jameswallacew8]

Hi Mark,

Thanks for your reply.

The centroiding algorithm currently returns a sum of the intensity and an intensity-weighted average of mz and ion mobility value for peaks that were deemed to be close together based on their fast mz integer value. Would it perhaps be better for isotope detection to select the peak in a centroid with the highest intensity and report the actual m/z, intensity and ion mobility values as opposed to this averaging and summing?

Thanks,
James

[markmipt]

Your suggestion is quite interesting. I'll add that functionality as biosaur2 optional parameter later

Mark