Page 1 of 1

PDA Peak Purity value inconsistent

Posted: Tue Oct 10, 2017 8:00 pm
by marcojfelix
I received this doubt from our lab about the Peak Purity calculation in Clarity, we are used to work with PDA peak Purity in EZChrom and Chemstation, but the guys in the lab are not understanding this concept in Clarity with our Azura PDA 2.1, see below and attached the case:

First setting, Purity Threshold = 990, Abs Threshold = 10% (same values we use in EZChrom PDA)
Any peak in the chromatogram is indicated as pure, even the peaks that I know are coeluting with something else (in the image is overlapping mix of impurities + placebo to demonstrate coelution) such as Impurity A and Impurity C

Second setting - Purity Threshold = 0, Abs Threshold = 10%
In this case, if we consider as pure only the peaks with PDA Peak Purity greater than 990, almost no peak is pure, only the less intense peaks, and we know those peaks showing no coeluting with placebo is pure.

If take in consideration our experience with EZChrom and ChemStation softwares and the theory of these settings, we believe the most useful setting should be 990 for Purity Threshold and 5-10% of Abs threshold, is that right for Clarity also, right ? Can we consider above 990 acceptable for a pure peak like we consider in other CDS ?
In this case why the peak purity value with PDA2.1+Clarity is not working ?
This is the first time we are using PDA peak purity with PDA2.1 + Clarity, since we have installed the new Azura systems here in July, we have not validated this tool yet, so we are not sure if this feature is working properly !

Re: PDA Peak Purity value inconsistent

Posted: Wed Oct 18, 2017 12:15 pm
by Daniel Mentlik
Dear Marco,
thank you for your inquiry. Your customer has found a problematic behavior in Clarity which was already repaired into version 7.3 when we were rechecking the peak purity calculations. The internal issue number under which we have tracked this issue was ISS-75594.

The problem with Peak Purity Threshold was that the spectra with lower similarity to clean spectra then the set value were discarded for the purpose of calculating the final Peak Purity. While setting the Peak Purity threshold excessively high, this might mean that only 1-2 spectra out of all spectra collected in the peak were used for calculation of final Peak Purity, marking the peak as clean (as obviously those 1-2 spectra had a similarity greater than the threshold, e.g. greater than 990/1000 in your customer's case). The tool was useful for setting up the discriminant value relatively low for peaks which had long ends and starts to eliminate the completely wild spectra from these areas, however for the same purpose the Absorbance Threshold may be used too. Setting high discriminant values completely voided the calculation.

As the Purity Threshold value might have been used for "cheating" the Peak Purity results in this way and there was no clear upper limit to set it to, we have removed the option to set the Purity Threshold form the software in version 7.3. I would strongly recommend setting the value to 0, notably if you are calculating the Absorbance (Intensity) threshold too.

For the Absorbance Threshold discrimination, the mechanism of calculation is effective "shortening" of the peak both at start and end for the purpose of the calculation of Peak Purity. A wavelength with highest intensity in the peak max spectra is found, and peak start and end for the Peak Purity calculation purposes is moved to a place where the peak height on the found wavelength is higher than the threshold for the first time (last time for peak end). Calculation is done on baseline reduced values, so in case peak start has intensity of 0.2, peak end 0.2 and peak apex 0.4, setting up a threshold of 10% means values higher than 0.22 (10% off 0.4-0.2). This also effectively means that in case you have "noisy" data, setting the threshold too low may mean the peak start and/or peak end may not be shifted by much and after the first/last point complying with the threshold a lot of points not complying witřh it may be present.

Final Peak Purity is calculated as arithmetic average of all valid spectra - depends on whether it is "All" or "5 Points" calculation.

I would strongly recommend to update to 7.3 or higher version, the Peak Purity calculations were repaired and some possible (but hard to invoke) crash locations were repaired in the version.

Best Regards,
Daniel

Re: PDA Peak Purity value inconsistent

Posted: Thu Dec 21, 2017 8:20 pm
by marcojfelix
tks

Re: PDA Peak Purity value inconsistent

Posted: Thu Dec 19, 2019 1:43 am
by marcojfelix
Hi Daniel,
After I have upgraded all the PDA systems with Clarity 7.2 to the new version 8.2 in one big customer, after some time all the users started to claim about the peak purity calculation compared to the old version, their criteria is >990 for specifity, but their settings used to be purity threshold 990 and abs threshold 10%, these are the same settings they are using in EZchrom and Chemstation, but the new versions of Clarity don´t have anymore the ability to modify the purity threshold.

I had to roll back all the stations to v.7.2 again, so that they can continue using purity Threshold of 990, after that the purity peak calculated values are equal again if compared to EZChrom and Chemstation.

Conclusion: I could not upgrade Clarity above 7.2 when the Customes Work with PDA peak purity because it is not possible to choose anymore another value for purity threshold fixed by DataApex around 200-300.

How can we solve that ? Is it not possible to bring back the option for purity threshold as all other CDS have ? (Enpower, Ezchrom, Chemstation).

Re: PDA Peak Purity value inconsistent

Posted: Fri Dec 20, 2019 5:22 pm
by Daniel Mentlik
Hi Marco,
we do not consider this option as anything that should be used, therefore it was removed. Please note that the Purity Threshold set as high as your customer does makes completely invalid results - let's assume the peak has 100 spectra total. Each of the spectra is compared to the apex spectrum to get kind of a match, ranging from 0 (no similarity) to 1000 (or 999, totally same spectrum). Then you start calculating peak purity for the whole peak as a means to define how much contaminants there are in the peak.

Setting Peak Purity Threshold to 990 means that for any of the spectra in the peak its match value is compared to the set threshold, if the match is lower then threshold, it is not used for peak purity calculation at all. From our hypothetical 100 spectra in the peak, very probably just the apex spectrum (its match with itself is 1000 or 999, depending on the software) and maybe one more have match higher then 990. In all this means that the Peak Purity (pureness of the peak) is calculated from 1-2% of data points of the peak and the resulting peak purity is always higher than 990, even if the peak is completely contaminated with impurities. In fact having completely impure peak will get higher peak purity in this case then completely pure peak (but still just in final range 990-999) because there will be only Apex spectrum which passes the peak purity threshold criteria.

I understand that it is nice having all results showing that the peak is completely pure, but these results are false, and are false the same in Clarity 7.2 or lower, EZChrom, Empower or Chemstation.

The Peak Purity Threshold was not meant to be set this high, it has very similar usage as Absorbance threshold to discriminate random noise spectra from the beginning and the end of the peak, however in such case it should have much much lower threshold settings to be effective. As it is not possible to define which is reasonable limit for the threshold, we decided in the end to remove the option altogether in order to not provide falsely positive results.

Please note that for version 8.3 a fix is planned for proper display of the Peak Purity Spectra View when Absorbance Threshold is used (ISS-93883).