COMPUTATIONAL MODELLING OF THE BONE MICROENVIRONMENT

The primary aim of this project will be the development of an agent-based model of the myeloma micoenvironment supported by clinical images and other datasets. Myeloma is an almost invariably incurable cancer of the bone marrow. The disease usually responds well to chemotherapy and enters a remission phase. During this phase, residual myeloma cells are thought to reside in specialist niche environments within the bone marrow which can nurture and protect dormant cells but subsequently stimulate these cells to rapidly grow, divide and lead to re-accumulation and relapse. Understanding the complex interplay between the cells, tissue and bone is essential for developing potential therapies to prevent relapse and improve patient survival. However, it is difficult to observe dynamic behaviour in the 'niche' in vivo, but emergent system behaviour and "what-if" scenarios can be explored using computational models.

This project would involve the development of an agent-based model representing the myeloma niche, where one single virtual cell, or "software agent" represents a biological cell, informed by biological data generated by biologists and clinicians working in the field. The model would be implemented using our highly optimised FLAME GPU simulation environment, and the efficiency further improved by the application of state of the art machine learning and model reduction techniques.

Aims and Objectives:
Develop an agent-based model of the normal bone microenvironment
Validate this against available biological data
Extend model to include disease characteristics
Translate model to highly optimised FLAMPE-GPU environment
Simulate disease outcomes and develop methodology to integrate with machine-learning techniques in order to develop simplified "surrogate models" of the myeloma development process.

The novel aspects of this project are i) development of a detailed agent-based model of the bone microenvironment and myeloma and ii) (Engineering specific) development of surrogate modeling process by integration of simulation and machine learning approaches using medical images and iii) (Engineering specific) specific addition of this functionality to the FLAME-GPU simulation environment.