One of the unique aspects of the SABS R3 programme are the software projects our students work on. Our cohort is split into three groups, who work on an open source project in each of our three research themes. The SABS R3 Research Software Engineer, Dr Martin Robinson, works closely with the students, helping to guide their work and reinforce the training received in the first module of the Programme.
The software projects for 2019-20 are:
- Pharmacokinetic-Pharmacodynamic (PK-PD) Modelling
- Supervisor: Dr Ken Wang, Roche
Utilization of mechanistic or semi-mechanistic mathematical model to quantify pharmacokinetics (PK) and pharmacodynamics (PD) have gained increasing importance in pharma industry throughout the discovery/development phases. The demand for PK and PK-PD modelling are high in pharma for both pre-clinical (discovery & translational) and clinical projects. Currently such “modelling” tasks requires skills in mathematical modelling (e.g. knowledge on model fitting, parameter estimation and uncertainty) and basic programming. The goal of this project is to develop a software package with a simple and user-friendly interface coupled with fast ODE solving and robust parameter estimation/ inference. This would be an ideal solution to promote PK-PD modelling to much wider community in pharma, enabling application of PKPD modelling to more complex projects (i.e. really moving away from the simplistic index or threshold PKPD approach).
- Expansion and Improvement of the Fragalysis Platform for Follow-Up Chemistry in Early Stage Drug-Discovery Projects
- Supervisors: Prof. Frank von Delft and Dr Rachel Skyner
The cloud-based Fragalysis (https://fragalysis.diamond.ac.uk) platform has the potential to lead the way as an integrated platform for early-stage drug discovery and should guide the user from fragment hits to more potent binders, quickly and cheaply. The code base is designed as open-source, so is deployable at other facilities/institutions for internal use and development and can serve as a repository for diverse computational chemistry algorithms from the community. The project is linked with the Eu Open Science Cloud and IRIS/ the Ada Lovelace Centre (ALC).
14/25 listed SABS partners (e.g. Exscientia, Oxford Drug design) research in areas where the project is scientifically applicable. The current incarnation of the platform allows the user to review protein-ligand structures from their screening campaign, view the protein-ligand complex with an indication of interactions in the binding site, and interact with ‘vectors’ representing an elaboration point on the molecule. These vectors query a graph-network and provide the user with collections of commercially available compounds that they can purchase for a follow-up screen. This project will develop a python API that will allow users to upload their own data and will provide access to the underlying algorithms in an open source fashion, with direct access to compound libraries and open-source screening project data such as the Target Enabling Package (TEP) data from the SGC.
- MRI Brain Segmentation Optimized for Elderly Brains for Use in Neurodegenerative Disease
- Supervisors: Dr Elisabeth Grecchi and Dr Christopher Buckley
Accurate anatomical segmentation of structure of interest is critical in the neurodegenerative disorder field. Quantitative analysis of medical images is often an important endpoint in research and clinical trials. Atlas-based MRI brain segmentation tools are widely available and used in research settings, however, brains of individuals suffering from Alzheimer’s Disease or Mild Cognitive Impairment are often characterised by localised or widespread atrophy, and pathologies are characterised by enlarged lateral ventricles when compared to a healthy population. Traditional atlas-based MRI segmentation tools have often fallen short in the segmentation of these specific regions, considering they rely on the original segmentation of healthy rather than atrophic brains. The goal of this project is to develop a brain MRI segmentation tool that provides accurate robust segmentation of problematic brain regions across the neurodegenerative spectrum. Furthermore, the methodology should be generalisable to perform well with the typical variance in MRI acquisition parameters and other factors that influence image contrast. The tool should be available and usable by the broad community without constraints due to necessary ancillary software or hardware. Accuracy and computation time should be comparable with commonly available methods.