Open AccessA mathematical framework for modelling 3D cell motility: applications to glioblastoma cell migration
Author(s) -
Michael B. Scott,
Kamila Żychaluk,
Rachel N. Bearon
Publication year - 2021
Publication title -
mathematical medicine and biology
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.542
H-Index - 45
eISSN - 1477-8602
pISSN - 1477-8599
DOI - 10.1093/imammb/dqab009
Subject(s) - workflow , computer science , in silico , glioblastoma , motility , set (abstract data type) , experimental data , data set , computational biology , artificial intelligence , data mining , biological system , biology , database , mathematics , cancer research , biochemistry , statistics , genetics , gene , programming language
The collection of 3D cell tracking data from live images of micro-tissues is a recent innovation made possible due to advances in imaging techniques. As such there is increased interest in studying cell motility in 3D in vitro model systems but a lack of rigorous methodology for analysing the resulting data sets. One such instance of the use of these in vitro models is in the study of cancerous tumours. Growing multicellular tumour spheroids in vitro allows for modelling of the tumour microenvironment and the study of tumour cell behaviours, such as migration, which improves understanding of these cells and in turn could potentially improve cancer treatments. In this paper, we present a workflow for the rigorous analysis of 3D cell tracking data, based on the persistent random walk model, but adaptable to other biologically informed mathematical models. We use statistical measures to assess the fit of the model to the motility data and to estimate model parameters and provide confidence intervals for those parameters, to allow for parametrization of the model taking correlation in the data into account. We use in silico simulations to validate the workflow in 3D before testing our method on cell tracking data taken from in vitro experiments on glioblastoma tumour cells, a brain cancer with a very poor prognosis. The presented approach is intended to be accessible to both modellers and experimentalists alike in that it provides tools for uncovering features of the data set that may suggest amendments to future experiments or modelling attempts.