Towards efficient drug development modelling with machine learning.

1) Brief description of the context of the research including potential impact:

Despite significant investment, a substantial proportion of Phase III drug trials fail, contributing to delays, rising costs and unmet medical needs. One of the key challenges is accurately predicting drug effects during development, particularly in pharmacokinetics (PK) and pharmacodynamics (PD) analyses. By integrating machine learning (ML) with PK/PD modeling, this project seeks to address these challenges.

Advanced data analytics, coupled with natural language processing (NLP) tools, will be used to improve drug development workflows. The potential impact includes reducing the time and cost required to bring new medications to market, improving success rates in drug trials and enhancing patient outcomes.

2) Aims and Objectives

The primary aim of this research is to develop innovative ML algorithms that can enhance drug effect predictions by integrating existing pharmacokinetic and pharmacodynamic knowledge with machine learning techniques. The specific objectives are:

- To create a comprehensive PK database that integrates open-access scientific literature and proprietary GSK data using an NLP pipeline (PKPDai).

- To develop and test ML algorithms that predict first-in-human pharmacokinetics by analyzing quantitative PK parameters and physicochemical properties. - To expand the PKPDai pipeline to include PD data, creating a robust PK/PD database for ML applications.

- To develop automated PK/PD modeling techniques that leverage prior knowledge and improve prediction accuracy using machine learning frameworks. 3) Novelty of Research Methodology:

This research stands out for its integration of machine learning with pharmacometric modeling, leveraging prior distributions and existing PK/PD data to enhance predictive accuracy. The use of NLP pipelines to systematically extract and compile PK/PD data from literature and internal sources is another innovative element. Additionally, the project incorporates nonlinear mixed effects modeling and ML techniques to analyze time series data, creating a more robust and scalable framework for PK/PD prediction.

This novel approach aims to outperform traditional pharmacometric models, particularly in predicting drug behavior in early-stage clinical trials. 4) Alignment to EPSRC's strategies and research areas

This project aligns with the EPSRC's focus on digital healthcare technologies and advanced data analytics. By applying ML to pharmacometrics, it contributes to the EPSRC's goals of using AI and data driven methods to enhance healthcare outcomes and streamline drug development. The project also supports the EPSRC's commitment to driving innovation in healthcare technologies, improving cost-efficiency, and enabling faster delivery of new treatments to patients.

Furthermore, the use of advanced ML techniques aligns with EPSRC's strategy of promoting data science and AI as transformative tools in scientific research. 5) Any companies or collaborators involved:

This research is conducted in collaboration with GSK, a global leader in pharmaceutical innovation. GSK will provide internal data and collaborate on developing the PKPDai pipeline and testing the machine learning algorithms on real world clinical data. This partnership ensures the project benefits from both academic research and industry expertise, fostering innovation that can be directly applied to drug development processes.