Efficient factorial designs and the analysis of macromolecular crystal growth conditions

Abstract

Statistically designed experiments for screening crystal growth conditions greatly reduce the amount of sample necessary to find conditions for crystal growth and lead naturally to a useful database for improving crystallization conditions in cases where the initial trials do not produce adequate results. Designed experiments simultaneously vary all the factors to be screened, and this results in a substantial increase in sampling efficiency. Essential features and examples of two-level factorial, fractional factorial, and generalized multilevel incomplete factorial designs are compared and evaluated. Rudimentary microscopic examination can be used to determine the experimental outcomes, which in turn can be used in subsequent optimization and analysis. Despite the limitations of this crude experimental assay, which include subjectivity, poor precision, and mistaken identification of microcrystalline for amorphous precipitates, quantitative evaluation of the ensemble of tests in an experimental design has revealed a number of significant aspects of crystal growth behavior both in screening and in more detailed explorations. Practical ways to improve the accuracy of quantitative evaluations are available, and these would substantially enhance the usefulness of designed experiments in finding and optimizing crystal growth conditions.