Every time the method is run it gives a non-linear response for an antibiotic compound (MW=555) on our Varian 1200L MS/MS. I can not even find a linear portion of the standard curve over more than 3 consecutive data points. A metabolite from the compound (minus a methyl group) and 2 other antibiotic compounds run simultaneously in the same method are all linear. We are using stable isotope dilution tandem mass spectrometry in the positive ESI mode. Do you think this can be published? I am concerned that the referees might not allow this even if it is scientifically justified. Should I go back and try different chromatographic conditions to see if I can get a linear response for all the analytes?
If you cant find any good explanation for your observations yes. Remove this compound and olny publish the ones with linear curves.
best regards
The compound elutes at 7.5 minutes and is well removed from the solvent front, there is a slight overlap with its metabolite (minus a methyl group) which elutes earlier and gives a nice linear response. The calibration or analytical range is from 5ng/ml to 8000ng/ml. The flow rate is constant at 0.2ml/min and the gradient runs linearly from 50% methanol to 87.5% methanol plus 10mM ammonium formate in 7.6 minutes. I am looking for guidance on if I should try and publish a nonlinear method for one of the compounds or go back and redevelop the method.
Possiblely this compond under your conditions did display a nonlinear curve per your solid experimental procedure. I think it can be published for scientific communication and discussion in the area. However, it is necessary for you to go back and redevelop new method to justify and confirm your previous results. Best wishes.
Can you make a curve fit, for example quadratic? If you can (and it is reproducible across runs) then why try to force it to be linear.
If your compound is fairly concentrated, the detector could be approaching saturation (EM tubes often give a non-linear response at the top end). If you have a weak solution you may be getting some ion suppression that is only significant at the bottom end. Either way, these are genuine and justifiable reasons for a quadratic fit.
Richard
Can you publish it? Sure. You can generally publish anything. Would you want to publish it? Maybe, maybe not. It sounds to me like your method may not be fully developed. I would suggest going back and working on your method again to resolve all of the compounds (you mentioned you had co-eluters). Deal with any matrix issues or sample problems early. It will be easier to support a non-linear result if the method quality is high. Some methods are non-linear (due to many things, such as use of detectors that do not output in linear fashion such as CAD or ELSD) and that is OK as long as your can show a repeatable and good quality curve fit. *If you can scientifically support it, it should be OK, but if the method is problematic, better to sort that out now.
A calibration curve does not have to be linear, but you must be able to validate the assay. If the signal appears to be saturating at high concentrations then the data may fit a hyperbola, but a quadratic may do. How reproducible is your curve? Can you get the required intra and interbatch RSD values that you require?
I would suggest that you reduce the concentration of your analyte, because the detector might get saturated and give you false results. Rerun the experiment after flushing your detector with a clean solvent and bake the column to remove any residues.
Hello!
Whether you can make public the method of this non-linear calibration curve, I do not know unfortunately.
But I suspect that the cause of the nonlinear behavior of the 10 mM ammonium formate is. You can change this if you, for example, Formic acid used. Whether but then you have to try out the chromatography sufficient.
Best regards
Burghard Cordes
If we are talking about linear range, we have to keep in mind that this is a pseudo linear part of ESI ionizatio. An absolut non linear calibration curve is not that unusual: e.g. surface effects and dimerization tendencies are concentration dependant and, hence, result in absolut non linear calibaration curves.
I saw accepted publication using small particle size columns and flow rates of 100 micro liters per min (if van Deemptet woud see that) and the peak at dead volumne was celebrated to show good retention behavior.
Thus, your issue seens to be a true behviour and if your method can be validated I do not ser any reason why not to publish - everybody dealing with this compound will be thankfull, as he/she does not need to invest any time to verify that non linear behavior.
I know it's a bit cross-disciplinary, but give 5-PL curve fitting a shot.
Information:
https://psg.hitachi-solutions.com/masterplex/blog/5-parameter-logistic-5pl-nonlinear-regression-model/
http://www.dartmouth.edu/~dartlab/uploads/Bio-RadTechNote2861_principles_of_curve_fitting.pdf
In bioassays, there are non-linear reaction rates that need to be compensated by non-linear curve fititng. In an MS situation, I'm using this to compensate matrix-binding effects. Not sure if that will fit your own implementation...
I am looking at the possibility of revalidating the method using different conditions, either on the HPLC or the mass spectrometer.
Agreed with Robin Whelpton that calibration curve does not have to be linear,
Do a linear transformation… the data does follow the curve ….conversely everyone wants one size to fit All and to accomplish everything by doing it all in One convenient run…that is not practical…. but if you split the assay into linear and non linear components and then dilute the high quantity compounds or samples you may stay in the linear range.p and avoid possible saturation or odd fit. I am dealing with the same issue in a 150 compound run.
- Badges
- Science topic
- Similar topics
- Instrumentation Engineering
- Instruments