292 — Predicting oral disintegrating tablet formulations by neural network techniques

I’m 80% done with #365papers as of today!

Oral disintigrating tablets (ODTs) are that kind of drug tablet that dissolves in your mouth without the need for water. Love ‘em. The old mint Alavert tabs were baller.

And ODTs were a huge success for the pediatric and geriatric population when they arose in the 90s due to the prevalence of dysphagia.

There are a variety of ways to manufacture ODTs, and the parameters to optimize the swallowability of the drug widely vary and interact in complicated ways. (Adding more filler increases the mass, but it also increases the surface area which improves dissolve-speed, for example.) A technique called SeDeM was developed in order to automate the balancing of these parameters, but it is limited in what parameters it can accomodate and it cannot account for dissolve-time of the ODT.

These researchers developed a neural net that — based upon pharmaceutical literature —

They used DeepLearning4j to develop a DNN which, given the formulae of ODTs, can predict disintegrating time, the release profile, and the physical stability of the tablet, with mid-80%s levels of accuracy.

This means that manufacturers can begin to rely increasingly upon automated QA techniques, rather than relying so heavily upon the intuitions and use-tests of pharmaceutical formulation experts.