Q:Can several target variables (or outputs) be considered at the same time in one process optimization project?
Yes. With DCT, we simultaneously address multiple target variables as long as they can be accurately measured and quantified. For example, we can include characteristics like cell toxicity, protein stability and antigen binding activity in an optimization attempt for antibody production.
Q:If I have multiple target variables, is the number of experiments going to increase linearly?
No. When a project contains multiple target variables, the number of experiments will increase because DCT dictates that the system be investigated from various different perspectives. However, the increase in the number of experiments is not linear. The number of experiments will depend upon the requirements that different target variables have for the system. If different target variables require similar system configurations – that is they have similar parameters – the number of experiments will not increase significantly. However, if different target variables require different system configurations with different parameters then more experiments will be required.
Q:Does DCT analyze the interaction among parameters at all? What about correlation among parameters?
No. When we design experiments using DCT, we don’t analyze the chemical or physical characteristics of each parameter, or the interaction between them at the physical or chemical level. We may do analysis for correlation among parameters at the system level. The purpose of doing this kind of analysis is to help us to understand each parameter’s contribution to the system. Our focus is to get the best results in the most efficient way. If the interaction between parameters is important to deliver the objective, then all the related parameters should remain in the system after the optimization. The optimized system provides the foundation for the scientists to do further research on the interaction among the parameters.
Q:For a system with two or more target variables, if the overall objective is to maximize one variable (e.g., recovery yield) while minimizing the other output variable (e.g., impurity level), and maximizing yield comes at the expense of a higher impurity level, how do you optimize such a system to achieve the ultimate goal of highest yield with lowest impurity?
It is a physical reality that higher yield often comes at the expense of lower purity regardless of methodology. Depending on the importance or priority of each target variable, an equilibrium point with a relatively high yield and a relatively low impurity level that can satisfy the business need should be defined. With DCT, only a few experiments are needed to, reach this equilibrium point.
Q:What kind of parameters are eliminated or remain in the optimization process?
When we use DCT to optimize any system, all necessary and important parameters remain in the system. Their contribution (or importance) is evaluated individually with sensitivity analysis. Usually, in a very complex system with condition parameters and many material parameters, condition parameters are determined to be more important than some of the material parameters and will remain in the optimized system.
Parameters like temperature and pH, vital in processes involving the growth of cells and bacteria, will never be eliminated in any design as long as the client wants us to include them in the optimization process. If certain parameters cannot be modified by the client , these parameters become restraints to the system and the solution we provide will be the most optimized experimental design possible under these restrained conditions. For experimental design purposes, we don’t need to know what those restrained conditions are, and we only optimize the parameters that are controllable.
It is possible that some common “condition” parameters such as oxygen supply, could be eliminated from the system. For example, oxygen could be harmful to the growth of an anaerobic organism. Sometimes, parameters that are firmly believed to be necessary based on experience may not be necessary at all when they are optimized within a comprehensive approach. For example, in our vaccine production case we eliminated two amino acids. Once they had been removed from the system production was increased by 300% and costs fell by 40%.
The elimination decision has nothing to do with the value range that the client suggests. It is driven solely by the contribution that each parameter makes to the system and it is entirely possible that the optimal value of a certain parameter is not in the range provided by the client. With DCT, we can use an extrapolation method to determine whether there is a problem with the value range of individual parameters and determine their optimal values. Under these circumstances, a few more experiments may be needed.
With DCT, we can optimize an existing system without eliminating any parameters, or we can develop a new system with fewer parameters. If a system with fewer parameters can achieve the same results as the one with more, the simpler system is the optimal one.
Q:Have you ever conducted experiments based on a client’s design which have failed completely without generating any numerical results?
No. We have never seen this happen. If all of our initial system diagnostic experiments failed, it would mean that the system itself had serious problems. Something would have to be very seriously wrong with either the parameters included in a system or value ranges for those parameters. If such an event were ever to occur we would work together with the client to reconfigure the system and help the client to identify which parameters are causing the problem.
Q:How do you incorporate the concept “Time Course” in the design?
Time (as in fermentation, incubation or culturing time) can be used as a parameter and be included in the experiment designs. If included, the best result will only happen at a certain “Time” under certain conditions. For example, an optimized fermentation process reveals that the best yield of a product can be obtained on the 6th day of the bacterial culture when other conditions meet our recommended values. Since many operation or logistic issues are involved in the production process, it is quite possible “time” becomes a restraint factor on the system. For example, the client may have to use exactly 14 days as the growth cycle and harvest the cells on the 14th day. DCT then provides the solution about how to obtain the highest
yield on the 14th day. We don’t need to know the yields from day 1 to day 13.
Q:Is the definition of independent/dependent variables involved in DCT?
No. When we design experiments using DCT, we don’t differentiate the variables by analyzing the chemical or physical characteristics of each parameter or the interactions between them. Our focus is to get the best result by optimizing the conditions and simplifying the system at the same time. With that achieved, scientists are able to conduct further research on the interaction among the parameters.
Q:Is experience needed to start a project?
Although DCT minimized the reliance on individual expertise, the success of DCT still depends on the integration of our technology and some expertise and knowledge from scientists and engineers. Before DCT is used to design the first 3 to 5 diagnostic experiments, the information we need includes the possible parameters that are related to the objectives, their value range, and the current level of parameter values and target variables if the scientists have previously performed experiments. Please note that these data do not have to come from the scientists’ direct experience for a particular project, they could come from publications, or from their expertise.
This is all the expertise we need to start a project. Compared with traditional methodologies, which require very in depth knowledge of a system, DCT requires much less expertise.