protein RPHPLC nonlinear dose response calibration curve..

[alicee!]

hi all.. when i plot a calibration curve for reverse phase method response.. i.e. amount injected vs. total area obtained for a protein.. the curve does not look linear.
ie. the slope for say 0.2-4% solution of 15 microgram injection (100%) is different from that of 10-120% of 15 micrograms. when plotted in the same curve ie. 0.2-120% the lower points near 0.2% show 200% recovery based on the amount calculation from the slope and intercept values obtained for this curve.

i tried plotting a log amount-log area curve.. and this fits well with desired accuracy.

is this common behavior of protein in such a method. is such behavior acceptable?

also in such cases how would %impurities be quantitated. would i have to inject 3-5 amounts, plot a log-log curve every time and calculate %impurities for the test sample/ is this common practice..
or is area% reporting okay.

also, how to report % purity.

thanks!

[Uwe Neue]

Could be carry-over. There are ways to check for this.

[alicee!]

hii.. it isnt due to carry over.. the blanks in between dont show anything..
thanks..

[danko]

Alecee,

I think we’ll need more information in order for us to understand the problem.

1. How do the peaks look like at the different levels? Do they get deformed at the higher levels?

2. At which level does the linearity disappear?

3. What is the R^2 by the way (less than 0.995)?

4. Also, would you clarify the sample load? Do you load 15 µg analyte, or 15 µL solution of a given concentration? Because it’s not clear.

5. What are the column dimensions?

6. Do you see a disproportional injection peak at dead volume, when the curve begins to bend?

7. Finally, if you load the same amount multiple times, do you see increase or decrease in the response/peak areas?

Best Regards

[alicee!]

1. How do the peaks look like at the different levels? Do they get deformed at the higher levels?
no they dont.. but till up to 4% solution i only see principal peak.. higher than that i start seeing the impurity peaks and so start doing a total integration.. my 100% is 15 micrograms/50 microlitres..

2. At which level does the linearity disappear?
0.2-4% shows a linear response.. 10-120% shows another linear repsonse..

3. What is the R^2 by the way (less than 0.995)? regression is okay.. but when i use the slope n intercept to calculate obtained amount for the response obtained.. i get 200-400% recovery for the lower amounts..

4. Also, would you clarify the sample load? Do you load 15 µg analyte, or 15 µL solution of a given concentration? Because it’s not clear. 15 µg analyte, 0.3mg/ml.. 50 mcl injection vol

5. What are the column dimensions? 2.1*150, 3 microns

6. Do you see a disproportional injection peak at dead volume, when the curve begins to bend? no i dont..

7. Finally, if you load the same amount multiple times, do you see increase or decrease in the response/peak areas? i see a consistent result..

is it very unusual to observe such beahavior in RP methods..?
n is it also unusual to use log-log curves to correct this?
thanks!

[danko]

Alicee,

First of all, sorry for misspelling your name/username in my previous post.
And now to the point: I think the problem lies in the injection volume. According to my calculations, you shouldn’t be injecting more than 35 µL volumes – and that, provided the sample solvent is identical to the mobile phase (the start composition if gradient, which I assume is the elution mode). If the solvent is different from the mobile phase and especially have a stronger elution property, then the maximum injection volume should be 10 times less – i.e. 3.5 µL in this particular case.
I don’t know what the solvent is, so I can’t deliver more concrete explanations. But you should be aware of the fact that (especially) in protein separations there are different chromatography mechanisms (within the RP- mode) and sometimes a different pH or salt concentration can contribute greatly to the elution strength of the mobile phase.
To make a long story short, I believe some of your compound is eluting at dead volume and that is particularly the case at the low concentration standard solutions. Which translates in higher recovery, given the samples have a different solvent/matrix composition.
Suggestion: Prepare more concentrated standard solutions ( 5 – 8 times)and inject correspondingly lower volumes.

Best Regards

[HW Mueller]

When looking into RP of Mab I didn´t see any nonlinear protein methods, theoretically I don´t see either why some should be nonlinear. Practically, one can imagine nonlinearity via the carryover mentioned by Uwe, the interference of system peaks as mentioned by Danko, but also by partial "permanent" adsorption on the column (would expect the lower peaks to be too low). Then there is also association or attachment to the vessel, etc., of proteins that could cause diverse problems. This is sort of an "academic" contribution, it could certainly be instructive to hear from some with more experience.

[Bryan Evans]

Have you looked at the apex for 120% peak?

[danko]

Hi Bryan,

The unrealistically high recovery at the low levels is an indication of disproportionate small peak areas corresponding to the lower calibrators/standards. So, I wouldn’t worry too much about the high level peaks.
Also, the curve shape – as I understood the case – was concave, which is another indication of problems at the lower levels.

Best Regards

[Bryan Evans]

Hey Danko -

I agree with your trouble shooting. And if a tiny bit is eluting in the void, it would be hard to detect (at the void).
And your theory agrees with concave curve type. And I understand that looking at apex at 120% (too see if detector is saturated)
doesn't agree with observed data.

Still - it takes less than 5 min to look up the data and look at the chromatogram. We're trying to trouble shoot without seeing any data.
Sometimes surprises do happen (at least I've found that to be the case).

[danko]

Bryan,

I don’t dismiss your suggestion at all. In fact, I would review all the data (especially the peaks) even if all the figures/results seem in perfect agreement with the expected outcome.
I just wanted to keep the focus on the lower levels, in case they didn’t seem too important.

Best Regards

[alicee!]

hi all..
thanks for the replies..
sorry if i have managed to miscommunicate this..

there is a 15 mcg loading/50 mcl.. fit for the column in use.. 4.5*150.. 3 micron..

my problem is.. which is pretty much similar with my sec and iec methods also.. that there is loss at the lower levels.. i.e. my higher range shows good recovery and linearity.. i.e. near 80-120% of 15 mcg..
but when i evaluate the recovery (in terms of area expected for say 1%, 0.5% of 15 mcg based on results obtained for 15 mcg).. the area values r much lower than expected. there is some protein loss at the column?

this is why when a plot a linear curve from 0.2%-120%.. for area response.. i get nonlinearity.. and the line is overpowered by the points near 100%.. showing incorrect accuracy at 0.2-1% solutions..

i wanna know if this is acorrect way to evaluate..
or shoud i be looking at only a lower range curve..
or fit log log..

this rp method is to be used for quantitation of impurites (near 0.2-1%) as well as for protein assay (near 100%).
this is actually a pharmacoepial method that i m using..
not understanding if this is a usual expected beahiviour..

thanks!!

[vickig]

it is not unusual to report linearity in two different ranges, especially if this is the way your method is used.

[alicee!]

it is not unusual to report linearity in two different ranges, especially if this is the way your method is used.

so in such a case..
1 using the lower range for impurities..
2 and higher range for total protein assay..

is that what you mean?
would 2 be okay to do?

also.. how to prove acccuracy in this case.
and how to report %impurities n % purity.

thanks!

[alicee!]

n is log log conversion of data not usually done?
thanks..