Peak Purity

[sga]

The Cs report at the end of the a run says:

"
The purity factor exceeds the calculated threshold limit.

Purity factor : 999.982 (70 of 151 spectra exceed the calculated threshold limit.)
Threshold : 999.992 (Calculated with 70 of 151 spectra)
Reference : Peak start and end spectra (integrated) ( ).

"

The fact that 70 out of 151 spectra exceeded the Threshold, implies that the remaining didn't exceed the calculated threshold ? Are these spectra then some how suspect ?

Another Question:
What is considered as an acceptable threshold or purity factor in a regulated environment.

[mattmullaney]

Well...I'm more familiar with Waters PDA software, so I'd guess that the 70 of the 151 spectra that exceeded the threshold are Actually the Problem Ones. (rather than the other way round)

In any case, it doesn't seem to me that this separation has achieved a peak in this case that is spectrally homogeneous.

The final question is one that I don't know, and probably your regulatory affairs dept. would have some insight into. I'd guess that all spectra across the peak should afford values less than that of the threshold...and that the purity factor be as close to 1000 as possible.

My thought is that, possibly, the peak you're working with is either very different than that of the standard you're comparing it to in concentration, or the standard is too concentrated, i.e. with a max absorbance greater than one, or that there is a problem with matrix matching.

[sga]

I think you are right.

When I inject a known pure substance, I get a response that say for example:

"Purity Factor is within the calculated threshold"

Now I Inject a sample I can not achieve this kind of a result, which essentially means that some thing else is co-eluting ?

What to Do ?

[mattmullaney]

Well...the separation conditions will have to be altered such that the co-eluter is moved away from your analyte of interest.

If your run is isocratic, try a gradient, or swap out ACN if you're using it for MeOH...

...and see what happens. Others may see a similar path forward.

Please, see what you think.

There is another possibility...if the peak of interest's standard chromatogram in your library is much different in concentration than the peak of interest's in the sample, errors could also crop up even if the peak in the sample chromatogram is either spectrally pure or free of co-eluting species, due to RI effects.

[sga]

Hi Matt,

Many thanks for your kind feedback. I tried what you suggested.


Its a bit perplexing to see a report that an entire peak is unpure.
May be I am doing some thing wrong ?

[kubowicz.tomasz]

Hello

I can see small "bump" between 3.1-3.2 min so I think there is co-elution .
You can try change options and overlay more spectra "slices" to see where you have mismatch.

Regards

Tomasz Kubowicz

[sga]

" you can try change options and overlay more spectra "slices" "

Where do I find this option ?

[mattmullaney]

Hi again,

I kind of agree with Tomasz. There seems to be at least three "co-eluters," one between elution time 3.1 - 3.2 min and two others at and around the tail of the peak of interest in the neighborhood of 3.6 min.

Further, at about 3.25 min and 3.41 min in the red-colored plot, the PDA algorithm seems to have found an additional two "co-eluters".

I don't know the software you're using, but there should be a way to pick a "clean" spectrum from one 3-D data "point" and overlay its spectrum with any of the "dirty" spectra to see how they differ. I wish I had a PDA in the lab in which I work now to show you an example using Waters Empower--doesn't matter much which software as they all (should) have this option.

Anyway...this seems to me to be a case of classic mis-match between the standard chromatogram and that of the sample. Is there a way, in the case of what you are doing, to spike the analyte that gives the peak of interest into the sample matrix, chromatograph this, and use it as the spectral library reference? Otherwise, it seems that you'd be looking into ways to "clean" the matrix or setting up an entirely different separation...or could the glassware you're using to make up the sample possibly be contaminated?

Doesn't seem that this is a case of an overly-concentrated library spectrum to me, and it's good at least to know that.

Please, see what you think. I'm sort of hoping for "dirty glassware" at this point.

[sga]

Thanks Matt and Tom for your efforts.

First of all : "Spectral library reference": I am not using a library at all.
Glass Ware: Is all new and mostly highest quality plastic, which we use only once.

There is co-elution or insufficient separation, that may be correct.

However what really confuses me is that the whole peak except for a single entry lies beyond the threshold. Appears to be beyond logic to me. When the algorithm averages 5 peaks from differnet time points and compares the areas, some of them should lie above this average.

Otherwise how could one reach this average value ? (I hope you get what I mean).

Now I have read the Spectra at the apex of the peak into a spectral lib and tried to do an spectral search on the edges of the peak. This looks as follows:

Peak is at 3.287 and I am comparing with the time point at 3.129. (before the so called bump).

The graphic appears to be almost perfect. What do you think ?

[mattmullaney]

I think we're at a terminology impasse:

It seems to me that you would to have chromatographed a standard solution containing your compound(s) of interest using the same elution program as the samples that contain your compound(s) of interest. The 3-D chromatogram of the standard solution's separation is what I refer to as the "spectral library reference."

Not necessarily co-elution...but spectral inhomogeneity. Remember...PDAs measure differences in UV-Vis spectra--not always will spectra differ just because of multiple substances eluting at similar times. Refractive Index effects, pH effects and matrix effects all can affect UV-Vis spectra.

Now, as to the overlay that looks quite nice...it is derived from one chromatographed peak of the compound of interest in a sample injection, is that correct? Okay. If so, and without the benefit of comparison to a "pure" standard, how can one be sure that the spectrum at the apex of that peak is either spectrally or chemically homogeneous? One case I can think of off-hand are PDA spectra of Schiff Bases such as the DNPH-derivatives of aldehydes, like valeraldehyde, for example. You'll get beautiful overlays of spectra...but there are multiple compounds making the spectra.

Lastly...I think I may read the diagram a bit differently than you do for the printout earlier. I'm not overly good with moving pictures around, so I'll describe what I can here. In the chromatogram you put up, there are two lines drawn at the peak base, one is dashed, the other is solid. The dashed line represents the threshold and the solid one represents the delta...difference in spectra from the threshold. Where the solid line is above the dashed line, these are the data points, the "70 of 151" that are above the threshold limit. The red section lower down shows the same thing, only turned "upside-down" and normalized.

I think I understand what you're getting at with the "mean," but I have to go now. In the meantime, someone else may be able to help...and if not, I'll do so when I have the chance.

Hang in there, SGA!

[kubowicz.tomasz]

Hello

I will send you some tips tomorrow...I think you need to adjust options in Chemstation.
You overlaid two spectra (SIG10132.D and Test) but it is not the way to check purity. You need to overlay spectra "slices" between start/end of your peak (at least 3-5). You could send me file and I could check it in my Chemstation.

Regards

Tomasz Kubowicz

[sga]

Hello Matt....thanks for your long write up was helpful. Will hang in !

Tom : Please tell me a mail id where I can send you the spectra.

Regards

[mattmullaney]

@ Tomasz, I'd also ,be interested in the tips for DAD with ChemStation. Good for me to know about other CDS and how they work. If you'd want, I could tell you a bit about Empower and the PDA.

@ SGA, to talk only about the "means," remember the idea of weighting. If you have a set of data points, say with numbers...for example 20, 21, 20, 19,...and then all of a sudden you get 80, 90, 150...and then back to 20, 22, 18, 19, 21...these high numbers get "averaged" in with the sets of numbers which are closer to each other in value within the larger set.

Of course, in the case of photodiode array detection, these aren't simply numbers, but close, they're vectors. A lot of really complicated math is involved in something that seems so simple...and deceptively simple at that.

[kubowicz.tomasz]

Hello

kubowicz.tomasz@gmail.com

Regards

Tom

[mattmullaney]

SGA...I think I see now...I must be reading that red diagram incorrectly. I'll keep my typing mouth shut a bit and let Tomasz have the reins.

I still think I'm reading the upper portion of the figure, where the chromatogram peak and the dashed and solid lines are shown.

More than Enough from Me for Now.