Do we have a demo data set that has functional data as both inputs and outputs?
@Peter_Hersh
@brady_brady
Hi Wendy,
as far I know not. I'm currently working with a customer on such a case, but this is under NDA. Chris is informed and we will try to get some simulated data at best out of this. However this will need some more time.
The key is that currently this is still a kind of manual process. You have to extract FPCs for both inputs and responses and relate them to each other through modeling, then you bring the models all together. Briefly said. The goal is to automate as much as possible in the future as the developers did with the golden curve analysis.
Martin
Hey Wendy,
Here is the table I have been using for this. It is a Glucose monitor and I look at the monitor pretreatment before 0 time point vs. after. The patients were monitored for 10 hours (600 minutes from the time column) before treatment and 240 minutes after treatment. It is a cool way to show off filtering and cleaning up the data as well. I drop the whole data table into FDE then apply a fliter and fit a model then save summaries for pretreatment and do the same with post treatment. For interpertability I would recommend limiting to 3 FPCs.
DO NOT SHARE this with customers.
Greetings Wendy,
I am sure that there are some pharmaceutical examples out in the ethos. An example that I can think of off-hand is the bead coating process for extended release and multi-active drugs. The process starts with a sugar sphere core that typically receives a drug coating, a seal coating, and an enteric coating. Coating takes a very long time in a fluid bed equipped with Wurster columns. There are many functional inputs such as the spray volume over time (starts with a small volume to get a thin coat on the sugar without dissolving it), a heavy volume mid process (to get coating on as quickly as possible), and ends with a light volume to create a consistent finish on the beads. Temperature and air volume are receipe contolled and will vary to compensate for changes in rate of spray and other concerns. I describe a 3-stage process since the same functional variables are in play for the drug, seal, and enteric coatings. There are many variations on this theme as equipment and processes differ across the industry.
Once the beads are coated, it is typical to have the analytical lab run tests over time such as dissolution. It is typical that at least two different dissolution tests are done; one in acid media (to ensure thads are resistant to stomach acid), the other in a basic media to determine the release profile in the intestine.
The models I have seen in the past have been poor since the fuctional inputs and dissolution outputs tend to utilize summaries or step summaries for analyses. It is no wonder that getting a bead coating process to be consistent so that dissolution targets can be maintained is thought of as somewhat of a black art by formulation scientists and phrama engineers.
I will try to flush out an example to play with as this question sparks my interest. You may find other examples such as high shear granulation functional inputs for powder formulas to meet compressability profiles.
I promise to share if I come across anything interesting.
Rob
