Chromatography Forum

This for the folks out there who have done some mathematical modeling as part of method development.

But, ironically, my question actually has to do with a case where you can't do modeling. So let me explain. We are currently looking at a situation where we can't use modeling, because there are too many small peaks, such that we can't track them (if we were setup on an LC/MS then it might be doable, but we just have the UV). What we want to evaluate is combinations of Methanol and ACN in mobile phase B. So my question is: Do you think it is reasonable to evaluate:

first: 25/75 ACN/MeOH
then: 75/25 ACN/MeOH

And to continue narrowing it down from there. Or is there a chance that 100% MeOH or 100% ACN (neither of which would be run, by this approach) might give something quite different, which we wouldn't know about unless we actually ran them.

Thanks for any suggestions.

Or is there a chance that 100% MeOH or 100% ACN (neither of which would be run, by this approach) might give something quite different, which we wouldn't know about unless we actually ran them.

That's quite possible.

I assume you are doing your screening with gradients and dealing with small molecules (as opposed to, say peptides), so taking a "brute force" approach to screening is definitely feasible. I you use a 150 x 4.6 mm column at 2 mL/min, you can get reasonable k* values for a full-range gradient in 20 minutes [if you have a 50 x 2.1 mm UHPL column, you can cut that down below 5 minutes]. With a 4-channel system, you could screen ACN, MeOH, 2:1, 1:1, and 1:2 in an afternoon, even allowing for equilibration time. Then you can play "count the peaks" to see which looks most promising. If you want to invest an overnight run, you could repeat at a different temperature.

Once you find the most promising, you can then "fine-tune" the gradient steepness, temperature, (and the ACN/MeOH ratio if you're so inclined). The various modeling programs help tremendously in the fine-tuning, but they are not absolutely necessary.

Hi hajdaei,

It is perfectly possible to model mathematically (empirically). If you use the right modelling tools, perform the right experiments, and model on the right data, there is no need to track peaks at all!

Fusion QbD has three features to address your situation:
1: It will generate an experimental design for you to execute automatically, based on the mobile phase composition ranges you wish to study, including the blending of different organic modifiers you specifically mention. In will even build these methods and sequence in your data system ready to run.
2: The analysis of "Trend Response" data allows for the effective characterization of the resultant "whole chromatogram", not individual peaks. This means there is no need to track peaks at all, only integrate them.
3: The automated "Best Answer" search will predict for you which mobile phase composition is best for your chosen goals.

Most importantly Fusion QbD can do this in conjunction with different column chemistries, aqueous pH, gradient time, and temperature to ensure all likely major effectors are evaluated together. This allows for proper characterization of all interaction effects. Only when characterised can they be exploited for the benefit of your method!

For more information take a look at my website here:
http://www.ciinformatics.co.uk/products ... -software/

Our contact me directly at bob.lott@ciinformatics.co.uk
Good Luck