A new powerful tool for method dev. and validation-Fusion AE

[Alfred88]

Dear Chromatographers:
I just attended a free workshop offered by Varian, Inc. on method development and validation, using the Quality-by-Design approach. Varian is selling a new software call Fusion AE (developed by S-matrix). This software may be the next revolution in chromatography.
If you are in a hurry, check out this page: http://www.smatrix.com/products.html
This software helps to automate the pre-screening, optimization, and validation. You just enter some minimal info, and the program will generate the completed sequence for you to run. The software will then optimize the method for you, using internal modeling methods. The output will be the parameters for a robust method (It will help define the region of robustness for a method). It can also run validation, and generates reports automatically. I never thought that all the difficult work could be reduced to a few simple steps! This new tool is much better than other currently available software (e.g. DryLab), because it is not using the OFAT (one factor at a time approach). It used full factorial design approach for method screening.
Highlight: It has helped Eli Lilly’s scientists to develop a method in one night (Previously, they could not develop a method for 2 APIs, and other impurities in 2 months).
Pro: Fast, efficient. Low cost ($5K). Works with hardware from Waters, Agilent, or Varian.
Con: Currently used with Varian’s Galaxie software only ($5K more). Non-linear optimization may not converge to an optimized solution. Works only for small molecules. Peak tracking is still imperfect. Needs some configuration to automate with 6 reservoirs and 5-6 columns.

Disclaimer: I don’t work for Varian or S-Matrix, don’t work in sale, and have not used this software. I just want to show gratitude to Varian for all the info.

[Uwe Neue]

My understanding is that you still need to do some homework up front. The program is strong in math and experimental design, but completely stupid as a chemist. If you tell it to run pH 4.5 with phosphate, it will obligingly do so and analyze the junk data that you get from such a stupid experiment.

In the hands of a well-trained chromatographer, it probably can be very good.

[Vlad Orlovsky]

SIELC has solution which cost $0 and is called Free Method Development, can't beat situation when somebody is doing work for you, but you still get your salary:
http://www.sielc.com/Services_MethodDevelopment.html

We usually spend no more than a day on any method development.

[aceto_81]

From my point of view:
I also think it's a strong math program.
One always need to keep an eye on the chemistry.
There are a lot of programs, which can do (full) factorial designs. (eg R-project for free)
When doing method development with the use of full factorial designs, one needs some objective parameters to measure, like critical resolution, a Quality factor, number of peaks, sum of resolution, ...
This software develops your method within a design space, which should be defined before you start your method development, and doesn't take into account the chromatography (like column length, ...).
While Drylab does take this into account.
Actually, when using drylab for a method scouting (2 gradients, 2 temperatures), you are doing a full factorial design for 2 factors.

I regulary use screening methods to quickly find the major factors contributing to the separation, but never thought about using a full factorial design to perform method development.

I think Fusion AE is a math package, where Varian has found a new use for.

regards

Ace

[sscala]

Any hints on literature about and theoretical basis behind full-factorial-disign-assisted HPLC-method development?

I would like to take a closer look at it (eg using R)!

Thanks!

[aceto_81]

Hi,

I have no literature at hand about the use of factorial designs for method development, but if you take a look at the articles about fusion AE, it should give you an idea about how things are setup:
Choose a response variable (how many peaks in the chroms, total resolution, ...) This variable should give you an idea about how your goals are met.
Then setup your design with tests.

To do things with R, you should take a look at the AlgDesign package for making full factorial designs, or D-optimal designs.
Then use the lm function to examine your results.

Here an example in R:
require(AlgDesign)
dat<-gen.factorial(2,3) #Generate design with 2 levels, 3 factors
results <- -5* dat$X1 + 0* dat$X2 + 5*dat$X3 + rnorm(8) #generates results based on the design, + noise
summary(lm(results~X1+X2+X3+X1:X2 + X1:X3 + X2:X3,dat))
#Here I only analyse factor effects and 2 factor interactions.


If you want to try with R, but have some questions about how to setup things, you can always contact me at aceto81 at gmail dot com

Best regards

Ace

[lmh]

You might also like to have a look at ACD-labs software. I've never used it (I trialled one of their MS packages, but for MS-related reasons didn't buy it), so I can't comment on how well it does things, but it claims to offer automated hplc method development based on both known data (test runs) and artificial intelligence guess-work about chemical structures (pKa estimation), and be able to suggest conditions that will give the most robust separation.

[Kimico]

Recent articles in LC-CG talked about it. I saw the presentation last year during a method validation conference in SF and I thought the the software was a good tool to aid in the method development process, however, it does not replace the separation scientist like Uwe mentioned. A book of reference "Tecniques for the automated optimization of HPLC separations" by John C. Berridge is also a good source if you want to do it yourself or need more information.