Development of an Agent Based Model to Study Combination Therapy to Overcome Antimicrobial Resistance in Gram Negative Bacteria

Project ID: 2531bd1702

(You will need this ID for your application)

Research Area(s): Driving value and security transforming health and healthcare
Digitisation and data
Artificial intelligence

UCL Lead department: School of Pharmacy

Department Website

Lead Supervisor: Joseph Standing

Project Summary:

Antimicrobial resistance in Gram negative bacteria constitutes a major global health threat. Understanding in vitro antimicrobial pharmacodynamics relies on simple time kill (static drug concentration) and occasionally dynamic experiments such as the hollow fibre infection model, where drug concentrations mimic in vivo behaviour. Bacterial colony forming units (CFUs) are manually counted periodically, sometimes on standard and drug containing plates (to study resistance development) and experiments typically last for 24-48 hours. These experiments are laborious and efforts to automate are limited by inability to reliably extract useful data (such as number of viable cells) or still requiring a number of manual steps. The methods are well established for studying individual antibiotics, and have been extended to study two-drug combinations, but are unsuited to combinations of three or more.

The mathematical analysis of such in vitro studies centres on ordinary differential equations to model growth rates with time and include drug effects with the Hill equation. These models assume the culture contains a homogeneous population and fits relationships to the average bacterial density. As such the stochastic emergence of resistance or complex interactions between drugs is poorly captured.

This PhD seeks to develop an Agent Based Model (ABM) which will aim to simulate in vitro antibiotic pharmacodynamic experiments, addressing the problems of low throughput in terms of experimental data generation and allowing for the simulation of complex behaviour including resistance development and the effect of antibiotic combinations. ABMs have been increasingly used in oncology with individual tumour cells modelled in time and space to predict optimal treatment combinations and scheduling. Through extracting and digitising literature data on bacterial growth dynamics, mutation rates and antibiotic activity, the model will include biologically-informed parameters to simulate in vitro bacterial dynamics. Model parameters will be calibrated with literature data and predictive performance tested against new experiments.