Believe to Achieve: Evaluating the Accuracy of the EZGC™ Chromatogram Modeler

As a new member of the lab, I have been working on expanding the libraries of experimental data that underly our EZGC chromatogram modeling tool. If you have never used the tool before I recommend you give it a try! It is a web based software program which, given a list of compounds, will recommend columns and run conditions that in many cases can effect their separation  and even produce model chromatograms.

One question that lingers for both customers and our internal users of EZGC is accuracy of the program’s recomendations and the chromatograms it produces. Is the agreement between the model and reality “good”, “pretty good”, “ok”? This is something we wanted to quantify for ourselves.

To test this, a list of 22 compounds was put together and entered into EZGC. We chose these compounds based on availability and some interesting features that occur in the model chromatogram as shown in figure 1.


Figure 1. EZGC output for the test list

Figure 1. EZGC output for the test list


The EZGC program recommends using a 30m x 0.25mm x 1.40µm Rtx-VMS column and predicts an  almost total coelution (Rs=0.55) of compounds 7 and 8, shown in figure 2,  as well as less than baseline resolution (Rs=0.88) between compounds 18 and 19, shown in figure 3. Zooming in on these areas (marked with red boxes in figure 1), we can see more clearly what is going on in the model produced.

Figure 2. EZGC predicted coelution of compounds 7 & 8

Figure 2. EZGC predicted coelution of compounds 7 & 8

Figure 3. EZGC predicted close elution between compounds 18 & 19

Figure 3. EZGC predicted close elution between compounds 18 & 19













Something to keep in mind when evaluating model chromatograms is that EZGC  assigns peak heights randomly to impart the chromatograms with  some realism. What this means is if you plug in the same compound list twice, the two chromatograms obtained may have different relative peak heights while predicted retention times, resolution values, and peak widths will remain consistent.

Now to answer the real question of how well does the EZGC prediction match a real chromatogram in terms of separating compounds of interest. Our goal is for EZGC to be rugged and accurate so that a user can Install an out-of-the-box column on any functional instrument, run under the reccomended conditions, and obtain a chromatogram with the predicted elution times and resolution. So that is exactly what I did!

Figure 4. Instrument 2Q


I installed a brand new in box 30m x 0.25mm x 1.40µm Rtx-VMS on our HP 6890 FID workhorse shown in figure 4. This is  one of our oldest instruments and one that I recently dissasembled to the bare bones and brought back into service. Nothing fancy here I can assure you.

By obtaining a CH4 dead time and using the EZGC Flow calculator I determined the true length of the installed column (see Do you measure the length of your GC column? For details), set the He flow rate appropriately to obtain the EZGC recommended 40 cm/sec linear velocity, and collected the chromatogram shown in figure 5.





Figure 5. Actual experimental chromatogram obtained with EZGC suggested conditions.

Figure 5. Actual experimental chromatogram obtained with EZGC suggested conditions.

fig 5 conditions 1


Surprise! Quite a striking resemblance to the model. Experimental and predicted retention times seem to agree, the predicted 7+8 coelution occurs, and the separation of 18 and 19 went better than predicted. It looks all well and good on the surface but let’s get analytical.

The following table compares predicted and experimental retention times. The experimental column refers to retention times (TR) and peak widths at ½ max obtained on the instrument while the EZGC column refers to the same values as predicted by the model.

EZGC table






















It looks like EZGC is quite accurate overall, with an average absolute error of 0.24 min across a 15 min chromatographic run. EZGC is designed to slightly overestimate peak widths and we can confirm from the data that this the case. This means the program generally errs on the side of caution by underestimating resolution, as apparent in the separation of peaks 18 & 19.

After putting so much work into collecting data for the EZGC libraries I am quite pleased to see the program functions as intended. The prediction is not perfect, but after all, few models are. Several factors contribute to the observable error including the quality of the library data, accuracy of experimental flow rate and oven temp programming, and broadening of experimental peaks.

The next time you need to choose a column or develop a method give EZGC a try! What comes out will give you a running start, saving you time, money, and sanity.

Tags: , , ,

Leave a Reply

Restek Domestic Customer Service



Your Full Name

Your Email

Company Name


Spam Block (Please leave this blank)

all fields required

Thank you

Your message has been sent. We will be in touch shortly.

Message not sent

Sorry, your message could not be sent at this time. Please try again later, or contact Restek or your local Restek representative via phone.