topic Re: Modeling of dissolution profiles in Discussions

Modeling of dissolution profiles

marta-simoes — Fri, 09 Jun 2023 23:36:22 GMT

Dear all,

How to fit kinetic models to dissolution profiles? The most common for pharma are referred to below. I tried the non-linear modeling with the weibull function and the linear (zero-order), but the rest is not available. Does anyone have a functioning script or a solution for this?

Many thanks in advance!

Marta

Re: Modeling of dissolution profiles

Mark_Bailey — Mon, 31 Aug 2020 21:11:05 GMT

Your description is brief, so perhaps you already know this answer and tried it. The Fit Curve platform provides a limited number of PK models to choose from. It makes such models easy to use and interpret. The Nonlinear platform is more powerful and flexible, but it requires more set up. You can define any one of these models as a 'custom model' for the Nonlinear platform. See this JMP Help entry for more information.

Re: Modeling of dissolution profiles

marta-simoes — Sat, 05 Sep 2020 16:59:32 GMT

Dear Mark,

Thanks for your feedback. The issue with the non-linear platform is that we need to define the parameters and provide them a value, exactly as described in JMP help you mentioned. But I don't know what these parameters are, they are meant to be determined by JMP (I was expecting so!). Depending on the values I provide for the parameters, the function may not work. For instance, in my jmp file the gompertz function in the non-linear platform is not working, but the Fit Curve provides a good fit! Same with Weibull function.

How can I solve this?

Many thanks for your support.

Marta

Re: Modeling of dissolution profiles

Mark_Bailey — Tue, 08 Sep 2020 13:01:24 GMT

The nonlinear regression in Fit Curve and Nonlinear both require starting values. Fit Curve uses heuristics associated with each model to obtain starting values from the data. Nonlinear does not have this feature. You must specify the starting values when you define the parameters in the Formula Editor.

The heuristics simply consider the interpretation of the parameter and estimate the starting value. For example, a parameter might represent the central tendency of the curve, so the mean or median are good choices. Or another formula might have an inflection point, so the mid-point might be a good choice.

The regression then searches for estimates (minimizes loss function) until any of the convergence criteria are met. It is not a closed-form exact solution.

Yes, some functions are particularly difficult and demand more careful selection of the starting values.

Re: Modeling of dissolution profiles

marta-simoes — Tue, 08 Sep 2020 15:32:48 GMT

Thanks Mark! It is clear now!

I was able to put all the equations working with some adjustments in those parameters.

Thanks for your help!

Marta