DryLab accuracy

The DryLab model utilized in Waters AMDS has additional requirements The number of sample components should not exceed 12 peak area% should be greater than 1%. These requirements are necessary to achieve greater prediction accuracy only. Any discrepancies could be corrected manually in DryLab using the data entry screen by manually entering the retention of the components from the scouting runs (to assign the peaks with a certain number). DryLab has been used for the method development of model drug candidates... 

HiPac (53) from Phase Separation is another commercially available software package. In several aspects this software is similar to DryLab, but its most important feature is that it can estimate the optimum mobile-phase conditions for the separation of the mixture at hand or only a selected number of peaks. Recently, ChromSword (commercially available from Merck, Germany) was introduced (34). It uses a retention mc el based on solvophobic theory. The input for this package can be the structural formulas of the solutes, the combination of structural formulas and retention data from a single run, or retention data from two runs. Data from additional runs are incorporated into the model, and prediction accuracy below 3% can be achieved under these circumstances. 

In the GC mode, two initial temperature-programmed runs conducted over the same temperature range (i.e., the same initial and final temperatures) at two different heating rates are required for initial input data. For the best combination of predictive range and accuracy, these rates should differ from one another by a factor of about 3. A measured value for and an average value of the column plate number must be entered as well. DryLab software then uses these input data to calibrate the model and consequently, simulating separations of other conditions. 

The accuracy of Drylab, based on the above relationships, has been analyzed in several publications [15-21], and comparisons of separation vs. DryLab predictions have been reported in dozens of other papers, many of which are summarized in Ref. [13]. In most cases, the accuracy of resolution predictions is of the order of 10%, which is usually more than adequate. Expressions such as Eqs. (2) and (3), which are suitable for predicting retention as a function of certain conditions, require only two experimental runs for DryLab implementation. Other variables, such as pH or buffer concentration, require three or more runs, in which cases a cubic spline fit to the data can be used for predictive purposes. 

With the exception of the coltunn, any one of the selectivity conditions in Table 1 can be modeled, or two of them can be modeled simultaneously, by DryLab (if the column is changed, new calibration runs must be carried out). The number of experimental runs required for DryLab simulation varies from two to nine [21], depending on which selectivity variables are chosen. When a wide range in some condition (e.g., pH) is to be covered, additional runs can be carried out for improved predictive accuracy [20]. Additional modes are available for normal-phase and ion-exchange HPLC, as well as GC. The user can also create virtually any mode desired. 

---