Bio-Modelling Seminar by Umberto Picchini
Plats: Ekologihuset, Maskrosen
Kontakt: sara [at] maths [dot] lth [dot] se
Spara händelsen till din kalender
The bio-modelling seminar series is a series of short interdisciplinary talks in mathematics and biology. The talks are 20-30 minutes long and then there will be time for questions and a discussion. Coffee/tea will be served.
Speaker: Umberto Picchini
Title: A two-compartments stochastic model of tumor growth in mice and tools for Bayesian inferenceAbstract:Pre-clinical cancer trials aim at understanding the dynamics of tumor growth and evaluate the e ffect of treatments such as radio- and chemotherapies in delaying this. A typical trial involves
repeated measurements of the volume of solid tumors grown in mice. Tumors are grown until a critical size is reached, in case of which the mouse must be sacri ficed for ethical reasons, or until
a planned end of study. These trials pose a statistical challenge due to the missing data caused by the sacri fice and due to the substantial variation in growth patterns between
subjects. Recently I have been involved with the modelling of such data in scenarios where their amount is scarce. Following a treatment, I consider the tumor volume in mice as represented by two unobserved compartments corresponding
to the fraction of the tumor which is killed by the treatment and the one that survives. These two volumes and their dynamics in the population of mice are then modelled using a hierarchical mixed-effects model. This is to understand the between-subjects variation, in addition to the intrinsic variation in individual dynamics. The latter variation is represented via stochastic differential equations. Statistical inference for the resulting model is challenging. I will briefly describe tools for parameter inference in a Bayesian setting and the main finding of the study.