Page 113 - Chiral Separation Techniques

3.7 Combinatorial Libraries of Selectors for HPLC 89

Since this method of screening initially operated by selecting groups of molecules
rather than individual compounds, and since the difference between both CSP 20 and
CSP 21 was small, it is indeed possible that our “best” CSP 22 was not actually the
most efficient selector of the original mixture. To confirm this, as well as to satisfy
our curiosity to uncover which other selector was very powerful, we prepared three
additional columns CSP 25–27 containing single proline-based selectors with
amines 26–28 as a control experiment. As expected from the previous work [8], CSP
26 prepared with amine 27 also exhibited a very high selectivity (α = 24.7 for (3,5-
dinitrobenzoyl)leucine diallylamide) similar to that of CSP 22. Surprisingly, CSP 24
and CSP 27, prepared with amines 26 and 28 respectively, afforded only modest α-
values of less than 4.
The rapid increase in the separation factors observed for the individual series of
columns reflected not only the improvement in the intrinsic selectivities of the indi-
vidual selectors but also the effect of increased loading with the most potent selec-
tor. Although the overall loading determined from nitrogen content remained virtu-
–1
ally constant at about 0.7 mmol g for all CSPs, the fractional loading of each selec-
tor increased as the number of selectors in the mixture decreased. Thus, the whole
method of building block selection and sublibrary synthesis can be also viewed as an
amplification process.
In the classical one-column-one-selector approach, the number of columns that
have to be tested equals the number of selectors. Using the chemistry described
above, this would require the preparation, packing, and testing of 36 CSPs. In con-
trast, our combinatorial scheme allowed us to obtain a highly selective CSP from the
same group of 36 selectors using only 11 columns (less than one-third). A simple
theoretical calculation reveals that the use of all 20 natural amino acids with 12
amines would lead to a library of 240 selectors. While the preparation and testing of
240 columns would be time consuming, a mixture of these selectors could be decon-
voluted using our approach with only 15 columns or just 1/16 of the total number of
columns that would otherwise be required. The parallelism advantage of the
“library-on-bead” approach with mixed selector column would be even more
impressive with much larger libraries of selectors for which the deconvolution by
splitting the library in each step to two or three sublibraries would rapidly lead to the
most selective CSP. Obviously, this approach can dramatically decrease the time
required for the development of novel CSPs.
Although the power of this combinatorial approach was clearly demonstrated, our
method also has some limitations. For example, in a hypothetical situation in which
only a single selector is active and all members of a much larger library are attached
to the beads in equal amounts, the percentage of the active selector in the mixture is
low. Despite its possibly of high specific selectivity (selectivity per unit of loading),
the actual selectivity of a mixed selector CSP may be rather small because of the low
loading of the specific selector. Accordingly, the peaks for both enantiomers may
elute close to each other and the actual separation may become impossible to
observe within the limits of experimental errors. Thus the sensitivity of the chro-
matographic screening may somewhat limit this approach. However, the number of
selectors that can be screened in a single column remains impressive.