chemometrics

[SRINIVASAN]

can u suggest anybody how to predict suitable method by using chemometric tools? Thanks in advance
srinivasan

[tom jupille]

Depends what you mean by "chemometric". There are a number of commercially available chromatography modeling programs available:
DryLab (Molnar)
ACD Chromatography Simulator (ACD Labs)
ChromSword (Isis)
Isis (Datalys)
Virtual Column (Dionex)

And there are a greater number of published non-commercial programs that do the same thing.

[bartjoosen]

Of course, you could develop your own model
But buying a software is most of the time easier if you aren't from IT, or have software programming as a hobby.

[SRINIVASAN]

i need develop an experimental design for hplc method using chemometric tools. i need to know about chemomatrics tools like PCA(principle component analysis)

[bartjoosen]

If you need PCA, experimental design, ANOVA, linear regression and all other statistical tools, you need definately statistical software.

You can write your own, with or without the aid of Excel, you can buy plugins for excel, you can buy Systat, Minitab, S-plus, and a few others, or you can download R.

Drylab and others mentioned by Tom is a chromatography modelling software. You put some runs in it, and it will predict the outcome of further experiments. This works pretty fine, but it isn't chemometric software in the sense you mentioned.

If all you need is a full factorial robustness test, you would probably get away with Excel and a good statistical book. Also you could try: http://www.webdoe.cc

If you really want to do PCA, PLS, MLR, and a few others, you need some software, but also knowledge about how to use it.

Good luck

Bart

[Mark Tracy]

ANOVA and PCA are usually used as post-experimental tools for data analysis. The results then can be fed back into the experimental design for the next iteration. I have read a fair number of articles where PCA was used to analyze data. Only once have I seen where PCA made the situation more clear, and numerous examples where it added confusion. Of course, one of the things that PCA does best is to reveal "con-fused" assumptions.

[bartjoosen]

I'm not sure why you need PCA, use the slopes of the regression lines, used for ANOVA. This will tell you what you need.

[promochrom]

You may use two approaches to optimize a method. One is factorial design. It screens the effects of relevant factors that may affect an analysis (such as pH, column temperature, buffer type, column type, etc).It helps to locate the most important factors and teh possible range. The other approach is simplex optimization that do stepwise test until you locate an optimum.

An effective way is using factorial design first. Once you locate the important factors, optimize their value using simplex design. Several years ago when systematic optimization of chromatography was hot topic, many researchers used such appraoch. You can find some nice tutorial articles from the journals during that period (1989-1995).

Although there are many software tool now, you may just go a bid deeper about the fundamental if you want to do it well.

Promochrom

[Koen Hollebekkers]

I also think computerized programs are much more effective when talking about method development.

But I would be interested in a good article by which tools like factorial designs are used.

[bartjoosen]

Koen,

For method development, programs like Drylab and Chromsword are wonderfull. If you had the chance to use this programs to develop methods, you won't do it the "old" way.


You ask for articles, but how would you apply factorial design? In robustness tests, or just screening design to see which factors are the most important to play with for method development, or others?
Or factorial designs more general?

Kind regards

Bart

[Koen Hollebekkers]

Hello bart,

Well actually I'm looking for articles related to method development. Despite DryLab is a great program, it's still expensive. For some labs some smarter calculations may work out as well.

[bartjoosen]

You can develop your own software in exel (like Tom Mizukami), which uses some underlying formulas of Drylab. Don't know if Tom is still on this forum, maybe he would share his document?

Or you can use some statistical program.
an example of this: http://cran.r-project.org/doc/Rnews/Rnews_2006-3.pdf page 2-7.

If you search on the web, you can find the structure of how the retention mechanism works. If you understand this, you can make your own optimiser.

I dont have articles at hand for the moment, but if I find some time, I will post some references.


Bart

[bartjoosen]

A quick link:
http://www.lcgceurope.com/lcgceurope/ar ... ?id=371552