4-Year PhD studentship: Towards efficient drug development modelling with machine learning

School of Life & Medical Sciences (B02), University College London

About us

A four-year PhD Studentship in Pharmacometrics and Machine Learning funded by the UKRI EPSRC and GSK is available within the Institute for Global Health. The studentship will commence from 1st October 2024 onwards, under the supervision of Dr Frank Kloprogge with subsidiary supervision from Prof Joseph Standing and Nuria Buil-Bruna.

Project Title: Towards efficient drug development modelling with machine learning.

Background:
A large proportion of expensive Phase III trials fail. In recent years Phase III failure has declined, in part due to the integration of model-informed decision making in earlier phases. Pharmacometric (pharmacokinetic/pharmacodynamic (PK-PD)) models are used at all stages of pre-clinical and clinical development, but they are based on mathematical and statistical principles dating from the 1970s. Developing these pharmacometric models remains a laborious task where highly qualified staff spend large amounts of time.

Aims:
The overarching aim is to enhance drug effect understanding through improved PK/PD predictions using Machine Learning (ML) and leveraging standardised and centralised big data. The distinctive feature is the integration of nonlinear mixed effects (or multi-level) modelling of time-series (repeated measure) data in combination with ML and prior distributions, something that has seen limited exploration and adds a novel perspective to the field of PK/PD modelling. Existing Natural Language Processing (NLP) pipelines, developed at UCL, will be used to collate PK/PD parameter prior distributions and ML guided PK/PD prediction algorithms developed in house will be further advanced to enable accommodation of prior distributions.

Timeline:
The successful candidate will use the first 12 months, i.e. MPhil phase, to conduct a detailed review of the literature to build hypotheses and to define research questions and corresponding objectives. This time should also be used for feasibility testing through data collection at UCL and GSK using readily available NLP pipelines and subsequent preliminary analyses. After a successful upgrade to PhD student status the successful candidate will use the remaining three years to develop ML algorithms that incorporate prior distributions for the development of PK/PD models on various types of time series data that arise across the drug development pipeline from pre-clinical stages to Phase III. Developed ML-based algorithms will be benchmarked against conventional methods PK/PD modelling methods.

https://www.ucl.ac.uk/work-at-ucl/search-ucl-jobs/details?nPostingId=9986&nPostingTargetId=23522&id=Q1KFK026203F3VBQBLO8M8M07&LG=UK&languageSelect=UK&mask=ext