Question about specific factor combination in screening DoE (e.g. DSD)

[#1]

Dear community,

I do a lot of protein buffer formulation screens with DoE and one important parameter is the buffer molecule (e.g. Phosphate, Tris, etc.) and its specific pH range.

Is there any possible design were I can test several buffer molecules in their pH range (e.g Histidin between pH 5.5 to pH 7 and Tris between pH 5 to pH 9) in one design experiment? The problem which I have is, that I cannot test combinations of buffer molecules because the pH of each buffer (e.g Tris at pH 9 and Histidin at pH 7) influences each other a lot in results in a new pH. 

Do I have to run a separate DoE screen for each buffer or is there a trick to measure all together and avoid global interefering of the pH of each buffer? Next to the buffer, other factors such as NaCl etc are also measured, which dont interfere with the pH so much.

Thanks for help

[#2 · Accepted solution]

I'm not 100% sure of the factor constraints you've explained but from what I can gather, can't you treat buffer molecule as a blocking factor, then use the linear constraints part of the JMP Custom Design platform to constrain the pH levels for specific buffer molecules? It's basically setting up something not too different than a separate 'DoE screen' as you suggest in your second to last sentence. It's just merged together in one design matrix, with appropriate model specification for the factors of interest. Including how to handle the blocking variable.

One other thought...Unless I'm missing something I don't think a DSD is a viable option because of the foldover nature of design wrt to the pH factor...you've got to have the pH levels fold over in the entire design. So a d-optimal design, with the appropriate model specification (including how to handle the blocking variable) is probably your best bet for a screening design.

I'm sure others might have something to suggest as well?

[#3 · Accepted solution]

By the way, if the design is easy to create as separate design for each buffer, it wouldn't be a big thing to concatenate both designs into one after table creation, and evaluate. Sometimes it is easier to think in simple parts and combine them, instead of building a more complex one. There are pro's and con's for each way.