Use this URL to cite or link to this record in EThOS:
Title: Input estimation in nonlinear dynamical systems for drug-discovery applications
Author: Trägårdh, Magnus
ISNI:       0000 0004 7223 9053
Awarding Body: University of Warwick
Current Institution: University of Warwick
Date of Award: 2017
Availability of Full Text:
Access from EThOS:
Access from Institution:
In mathematical modelling for drug discovery, nonparametric methods are an alternative to the more commonly used parametric methods, and have the advantage of requiring fewer modelling assumptions. This thesis considers nonparametric methods for performing input estimation (deconvolution) -- inferring the input to a dynamical system based on measurements of the system’s state. A typical application is to determine the absorption profile of an orally administered drug. Commonly used input-estimation methods are restricted to system models that are linear. This thesis aims to develop and evaluate methods which can be applied to nonlinear systems, and which are additionally able to provide uncertainty estimates. An input-estimation method is considered to be a particular choice of 1) prior, 2) function parameterisation, 3) desired statistical quantity, and 4) estimation algorithm. Two classes of methods have been selected and implemented: direct optimal-control methods and Markov chain Monte Carlo (MCMC) methods. These have been evaluated on two pharmacokinetic and two body-weight modelling applications, using simulated as well as real data. Evaluation was based on several criteria, including accuracy, computational speed, and usability. The results show that the methods can achieve good accuracy, provided that data are relatively densely sampled. Properly applied, optimal-control methods can achieve very high speed, approximately 0.1s for typical problems, at the expense of not providing uncertainty estimates. For MCMC methods, the performance is highly dependent on the method settings as well as on the problem. In many cases, MCMC running times can be significantly reduced by a suitable choice of function parameterisation and sampling method. In all cases, estimation is based on clearly stated, quantifiable assumptions.
Supervisor: Not available Sponsor: Seventh Framework Programme (European Commission)
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available
Keywords: RM Therapeutics. Pharmacology