|
Abstract:
|
Pharmaceutical researchers are continually searching for techniques to improve both processes and patient outcomes. An area of recent interest is the potential for machine learning applications within pharmacology. One such application is the unsupervised clustering of pharmacokinetic (PK) profiles, by considering a PK profile as a time series object. We review hierarchical clustering within the context of clustering PK profiles, and find it to be effective at clustering PK profiles and informative via the resulting data visualization of a dendrogram. We scrutinize measures of dissimilarity between time series objects to identify Euclidean distance for clustering PK profiles. We further find dynamic time warping, Fréchet, and structure-based measures of dissimilarity, then use subject level data from a new drug application as a case study to demonstrate how unsupervised clustering of PK profiles can be useful in helping to identify genetic factors known to cause inter-individual differences in pharmacokinetics. Such information may be used in the determination of individualized therapeutic strategies or appropriate dosages, known more broadly as precision medicine.
|
