Deconvolution mass spectra

[#1]

I am a JMP 17 Pro user and I am pretty sure what I want to do is pretty straightforward but I'm just not quite getting there. The problem is explained in the attached powerpoint where the figures help clarify the problem I am trying to solve. In this analysis, a single mass peak is obtained by scanning a voltage to deflect a beam of ions of specific mass through a slit into a detector that measures their intensity. The diameter of the beam is less than the width of a slit, so a single mass peak looks like a step function with a flat top. The width of the peak effectively represents the width of the slit. Now fast forward to real life where we can have 3 or 4 peaks at separated by only .05 or less atomic mass units. The measured intensity is then a convolution of these, and it would be tremendously useful to deconvolve and get the centroid mass value of each peak. I first tried the simple way - Distribution and fit continuous, but the type of fits available are not really suitable. I also tried "Fit Curve" to see if that might help but again no luck. I'm now thinking if wavelet analysis in FDE is the way to go, as a Haar wavelet seems quite appropriate, but I am struggling with the interpretation of the results (or maybe have not launched the platform properly). Realistically in the attached data, it looks like there is probably 4-5 peaks there. Thanks so much for any and all help! Best, Greg

[#2]

I should also add we usually look at these spectra with the y-axis (counts) on a log scale, which then clearly shows the presence of more than two peaks. We typically have very little electronic noise.

[#3]

I suppose you've already found and read through HELP)) Signal deconvolution using JMP ?

[#4]

Thank you for the reply! Yes, I had seen that post. I really like the simplicity of using the distribution but the drawback is it is limited to 3 mixed normal distributions. The fact that the peak shape is not described by a normal distribution but a step function is also not ideal, but I think I could live with that. I was trying yesterday with PLS and FDE but really didn't get too far. But today is a new day and we persevere...

[#5]

I am sorry I couldn't help. Now that I think of it, I read  this paper by Mowbray et al. recently. Maybe your solution or at least the path to it starts with what is shown in figure 20? If I understand this correctly, functional PCA should allow you to find the components necessary to deconvolute a mass spectrum. However, I don't think it is possible to define the number of components you want in the mix but, rather, functional PCA will show you how many components are necessary to recreate a mass spectrum (I may be wrong, though). If Mowbray et al. are only remotely relevant, than maybe Arzac et al. are of interest too.

Good luck on the mass spec front!