search

mobiusklein / ms_peak_picker

A small library to provide peak picking for software processing mass spectrometry data
20 stars 9 forks source link

lorenztian typo #5

Closed Kawue closed 5 years ago

Kawue commented 5 years ago

Thanks for this nice piece of code. There is a small typo in your peak_picker class. Its consistently wrong, so it does not matter but your documentations states it as "lorentzian". Would be nice to adjust either the code or the comments.

mobiusklein commented 5 years ago

Thank you for catching that.

I'll fix the spelling in the code itself, but the old misspelling will still work when selecting fit type in pick_peaks.

I've found the Lorentzian fit didn't find the centroid very well when dealing with clipped peaks, which happen fairly frequently when operating the mass spectrometer for throughput.

mobiusklein commented 5 years ago

I've made the name changes. Both "lorentzian" and "lorenztian" will work with fit_type. I noticed I've disabled the C implementation of the Lorentzian least squares, and it appears to be a long standing setting. I've re-enabled them after some manual testing on some "clean" data sets. If I remember correctly, there were intermittent issues with the full width at half max calculation it required early in this library's development, and I couldn't demonstrate it was objectively better than the simpler quadratic fit.

Theoretically, an FTICR's peak shape should be closer to Lorentzian.

image

image

But my experience has been that Orbitrap peaks tend to be more Gaussian.

image