Design and optimisation in organic synthesis: Methodologies for the optimisation of multi-step one-pot batch processes

Assume tha a three step one-pot reaction is tt be optimised: Lay out a design to map the first step. Depending on how detailed information is desired, this can be a fractional factorial, factoral, D-optimal, or a response surface design for quardratic m odels. Run the experiments and detrmine the responses. The response matrix Y = [y1 y2...yr] summarises the results. A principal component analysis of Y will determine the number of independent event yielding Y, i.e the number of independent reaction that are responsible for the products (desired product and parasite products). By fitting regression models to the score vectors from the PCA of Y it is possible to identify the variable that are significant and necessary to control for favouring the dessired reaction and, equally important, which variable shoul be controlled to suppress the paraste reactions. We can, of course, also use PLS to model the fisrt step and use the PLS(X) scores as extra variables for the second step Now we know the important vari ables for the first step. These are now added as extra design variables when the experimental design for the next step is laid out and we determine the the composition of the reaction mixture after the second step. The outcome of the first step defines th e input for the second step. By perturbing the important variables of the first step we will induce variation in the input for the second step. This is necessary since certain by-products may be detrimental in subsequent steps. By this procedure, we conca tenate teh first and the seciond step of the process. The procedure is iterated for the third (an for all subsequent) step(s). This technique makes it possible to establish quantitative models for multi-sstep batch pocesses. The model pays attantionto all possible interfering side-product and gives a possibilty to control the process. This will be important when the TAC regulation is implemented and that pharmaceutial production must account for possible variation i