Has anyone evaluate the chromsword software for HPLC method development?
Could you give me some reasons why you think it is useful or not.

Just a couple of quick opinions.

I think all of the structure based features are useless from a method development perspective.

I have never used the auto feature connected to Hitachi instruments.

Linear optimization of slovent strenght, temp, etc. is useful but nothing that you can't do with a spreadsheet or even better, a piece of graph paper. pH prediction of seperations tend to be error prone and off by enough to make the feature not very useful.

If I was going to buy chromatography simulation software I think I would buy the DryLab package.

Tom,
based on what you said, what's difference for DryLab? i beleieve it also uses same principles.

I'll jump in here with "unbiased" comments (by way of "truth in advertising": one of my jobs is DryLab tech support):

I actually partly agree with Tom M's comment: most of what DryLab (and other chromatography modeling programs) do can be done with a spreadsheet or a piece of graph paper -- assuming you know and understand the appropriate relationships.

However, some things (like modeling isocratic separations from gradient experiments, or 2-variable models) are "non-trivial", and some "bells and whistles" (like drawing simulated chromatograms and updating them in real time) provide very useful qualitative insight into what's happening to your separation.

And if you are doing more than a "quick and dirty" model (i.e., if you want others to be able to use it), you have to consider the cost of your time in setting up and documenting the results.

If anyone has specific questions, I'd be glad to provide the answers, either here in the Forum or by private e-mail.

By the way, you can download an evaluation version of DryLab from our web site:
http://www.lcresources.com/software/software.html

-- Tom Jupille / LC Resources Inc.

Actually, I thought that Chromsword and Drylab worked on quite different principles. Chromsword accepts inputs of structural formulae and predicts separations solely on this basis. However, I thought Drylab took the results of a few experiments, and then extrapolated these to cover all of the parameter space.
Thinking again, I had the impression that newer versions of Chromsword allowed some input of experimental data to correct predictions?
Have I got this all wrong?
Tom, I think you could usefully clarify at least half of this argument.

David

LOL! You're correct on the DryLab part. What it does is to use the results from a limited number of experiments (typically 2 or 3 runs per variable) to calibrate an appropriate mathematical model. Over the 15+ years that the program has been out, it has evolved quite a bit to include more (and more sophisticated) models.

We have, over the years, looked very closely at incorporating structural information, but we haven't found anything that's both general enough and accurate enough to incorporate into the product, especially since we can obtain the same information from a single experimental run (20 minutes or so).

-- Tom Jupille / LC Resources Inc.

Anon 1 - Customer service. I have brought in LC Resources for on site training and they were very responsive to our needs. They sponsor this discussion board, and have answered every direct question I have posed to them.

David - I'm not sure who you were directing your question to but I agree with what Tom J. was alluding to. In all of the software I have tried sturcture only based prediction is not accurate enough to be useful in developing methods.

Software like ChromSword that can use both structure and injections tend to abandon the structure model once data from real injections is entered. "Theory guides, experiment decides." So why not just start with the experimental data, it would probably take longer to enter all of the structure data than it would to make the injection.

My only real criticism of most of these chromatography simulator software packages is that they tend to hide what is actually being modeled behind the "bells and whistles". Optimization of %organic for an isocratic separation is probably modeling the linear relationship between log(k) and the orgainic fraction for each analyte. It would be great if the software provided access to (graph) this fundamental data instead of only the derived simulated chromatogram and resolution map. Similarly for gradient methods. I have not used recent releases of any of the simulator software so my opinions are probably getting dated.

http://www.lcgceurope.com/lcgceurope/data/articlestandard/lcgceurope/512001/5427/article.pdf

Tom, thanks for the kind words! You're right as far as the software "hiding" the basic models; the idea was that most users don't want to get involved in the details. In fact, the models used *are* documented both in the program itself and in the manuals for those who want to look "under the hood". You can even tweak things to generate "custom" models.

You're also right about the basic reversed-phase model being log(k) vs %B. As far as graphing the relationships, the latest version includes a spreadsheet "add-on" that does just that for reversed phase %B. If anyone's interested, we can provide a similar spreadsheet for any of the other models. We've also added some capabilities that actually *are* hard to duplicate manually, like finding the best 2- or 3-segment gradient profile.

As you implied, this type of software is a "work in progress" that evolves to take advantage of improved computational capability and (hopefully!) better meet users' needs.

-- Tom Jupille / LC Resources Inc.