Designing Diverse and Focused Combinatorial Libraries of Synthetic Polymers
Charles H. Reynolds · Journal of Combinatorial Chemistry · 1999
Molecular topology and genetic algorithm optimized quantitative structure−property relationships (QSPR) have been used to design diverse and focused libraries of synthetic biodegradable polymers. A diverse subset (17 polymers) of a 112-member virtual polymer library was selected based on the molecular topology of the repeat unit using a stochastic diversity method (SimSearch-SCA). These 17 polymers were shown to be highly representative of the two-dimensional property space for the full library where the properties of interest are glass transition temperature (Tg) and hydrophobicity as measured by the air−water contact angle (CA). The 17 polymers in the diverse library were used to derive QSPR equations for Tg and CA by using a genetic algorithm to select molecular topology descriptors for linear regression. High quality models were derived for both Tg and CA. These QSPR models were tested by comparing the computed and experimental Tg and CA values for the 95 polymers that were not included in the training set. Representative models give r 2 values of 0.89 and 0.92 for Tg and CA, respectively. The QSPR models were further tested by using them to build focused libraries with specific values of Tg and CA. The focused libraries were very successful in identifying polymers that fall within specified ranges of Tg and CA. This work illustrates that the same concepts of molecular similarity and diversity that have been exploited so effectively in the pursuit of small biologically active molecules can also be employed in the design of synthetic polymers, particularly in the context of parallel synthesis.