STATISTICAL-ANALYSIS AND MODELING OF CRYSTALLIZATION OUTCOMES

This work describes a novel application of a recently developed statis tical partial least-squares regression technique for the problem of es tablishing relationships between experimental variables in crystalliza tion trials and the experimental results. To validate this method publ ished sets of factorially designed crystallization trials were analyze d and it was discovered that these derived models show very good predi ctivity. These mathematical constructs cannot explain the detailed mec hanism of crystallization, but are a pragmatic and powerful tool which enables the crystal growers to set up crystallization trials not only in a rational manner but also with confidence. This is a useful and a general methodological approach particularly when crystallizing prote ins of limited supply.