The purpose of this software is to predict HPLC retention very accurately...
...subject to the following constraints:
There are two key features of this retention prediction system that distinguish it from others: 1) it makes use of an interpolated isocratic retention vs. solvent composition database and 2) it uses a novel "back-calculation" approach to precisely and easily account for differences between HPLC instruments.
1) Interpolated isocratic retention vs. solvent composition database
The heart of this retention prediction system is a database containing isocratic retention vs. solvent composition relationships of each solute. Data points are spaced closely together (at 10% B intervals) and interpolated to precisely determine retention as a function of solvent composition. This is absolutely critical for accurate retention prediction under varying gradient and flow rate conditions (and to a lesser degree, even between different HPLC instruments, see above).
In this demonstration version of the retention prediction software, the retention database holds data that was measured for 35 solutes. These 35 solutes were specially selected for demonstration purposes because they represent all of the 5 types of interaction that control retention on reversed-phase alkylsilica columns.
Of course, the retention data in this database is highly dependent on the specific stationary phase used, the temperature of the column during the run, the mobile phase solvent A and B, the mobile phase buffer, and the mobile phase pH. That's why we've defined a set of standard conditions, which are the same as the conditions we used to measure the isocratic retention data in our database. The standard conditions we used for the existing database are:
These standard conditions must be used in order for the software to be able to accurately predict retention. However, in the future, we plan to explore the use of other standard conditions that may further improve retention prediction accuracy.
2) "Back-calculation" of gradient and flow rate profiles
Armed with the isocratic retention vs. solvent composition database, you could theoretically predict gradient retention on virtually any HPLC system under a range of gradient programs and flow rates so long as you know two additional important pieces of information:
1) The precise gradient profile (solvent composition vs. time) actually produced by your HPLC
2) The precise flow rate profile (flow rate vs. time) actually produced by your HPLC
Together, these profiles fully define the output of the HPLC instrument (and therefore fully characterize differences between HPLC instruments).
In our previous work, we attempted to directly measure the gradient profile as accurately as possible for just this purpose. Unfortunately, it was an extraordinarily tedious process and the profile was very difficult to measure accurately. Nevertheless, we were able to significantly improve retention prediction accuracy this way. But it was clear that the accuracy (or lack of accuracy) in our profile measurements was limiting accuracy in our retention predictions.
To circumvent this problem, we devised a system in which the gradient and flow rate profiles aren't measured directly, but instead are back-calculated from a) the gradient retention times of a set of standard solutes (which we call the "gradient calibration solutes") and b) their isocratic retention vs. solvent composition relationships. Since we provide the isocratic retention vs. solvent composition relationships, the only pieces of information a user must provide are the retention times of the gradient calibration solutes in their run. From those, effective gradient and flow rate profiles may be back-calculated and used to accurately predict retention of any other compound in the retention database. In other words, the user-supplied retention times of a set of standard compounds are used to back-calculate what the gradient and flow rate profiles must have been to give those retention times.
The following figure shows a flow chart of this retention prediction process:
This retention prediction tool is fully functional, but in order for it to be at all useful to you, we need to expand the retention database. Right now, it only contains 35 compounds. The 35 compounds fully represent all of the 5 types of interactions that control retention on reversed-phase alkylsilica columns, making them a good set of solutes to test the system, but they aren't useful for any practical purpose because there are so few of them.
Assuming we had more compounds in the database, which we plan to add soon, you would need to do the following before you inject your sample:
1) Spike the sample with gradient calibration solutes
The gradient calibration solutes must be compounds with known retention vs. solvent composition relationships under our standard conditions. In other words, they must be in the retention database. There is nothing else special about the gradient calibration solutes except that they must elute over a wide range of retention times and they should elute over evenly-spaced intervals.
Of the 35 compounds in the retention database, we chose 1) adenosine, 2) N,N-dimethylacetamide, 3) p-toluenesulfonic acid, 4) N,N-diethylacetamide, 5) indole-3-acetic acid, 6) dimethyl phthalate, 7) indole, 8) diethyl phthalate, 9) diallyl phthalate, 10) di-n-propyl phthalate, 11) di-n-butyl phthalate, 12) di-n-pentyl phthalate, 13) di-n-hexyl phthalate, 14) di-n-heptyl phthalate, and 15) di-n-octyl phthalate
We chose to use 15 compounds because they allowed accurate back-calculation. If your gradient or flow profile has more features to capture, it would be wise to use more calibration solutes. On the other hand, if your gradient and flow profiles have less features to capture, you could go with less, but the minimum number of gradient calibration solutes allowed is 6.
2) Set up the standard chromatographic conditions
In order for the retention vs. solvent composition relationships in the database to be accurate, the standard chromatographic conditions must be strictly adhered to:
a) You must use one of the standard columns on which the database data was measured (right now there's only one).
b) Solvent A must be pH 2.80 30 mM ammonium formate buffer in water and solvent B must be 100% acetonitrile
c) The column temperature must be kept at 35 °C
One big advantage of this retention prediction system is that these standard conditions are relatively unrestrictive. You may change the gradient program and flow rate however you like!
However, one note of caution about flow rates:
The isocratic retention data was measured on the standard column at a flow rate of 200 μL/min. We tested the retention prediction accuracy of the system at flow rates of 100 μL/min, 200 μL/min, and 400 μL/min, and found virtually no difference in the accuracy of retention predictions. Still, one could imagine that viscous frictional heating caused by the mobile phase moving through the stationary phase could significantly change the retention behavior of compounds. Therefore, changing the flow rate more dramatically than we did could reduce retention prediction accuracy, but that's still something that must be tested.
Once you've run your sample under the standard conditions and determined the retention times of the gradient calibration solutes, you can begin using this software to predict retention of any other compound in the database.
Go through the step-by-step tutorial to see first-hand how gradient and flow rate profiles can be back-calculated and then applied to predict the retention of other compounds.