Master of

Computational Science Projects

Modelling of Lymph Node Structure and its Influence on Efficiency of the Immune System

== Background == The immune system is a complex system involving specialized cells exchanging information, intricate migratory patterns, and highly evolved biological data processing and storage systems. To detect foreign invaders the immune system detects specific biological molecules. A repertoire of t cells that each respond to different molecules are used to detect pathogens that are foreign to the body. The cells making up this repertoire reside in lymph nodes. Specialized immune cells patrol the body and find diseased cells. Once found they take up some material identifying the disease, so-called antigen. They then migrate to lymph nodes and try to activate t cells specific to the antigen being presented. This presentation process is mediated by a network of connective tissue in lymph nodes. T cells will stick to this network, as will the antigen presenting cells, seemingly using this network as a road system. It is speculated that this road network is evolved to improve efficiency of the antigen presentation process. If this is true is still an open research question however. Does the specific shape of the network actually have an influence on the efficiency of the antigen presentation process? If so, to what extent? And if the shape of the network is not evolutionarily adapted for antigen presentation efficiency, what alternate hypothesis for its specific shape can be proposed? Answers to these questions would give fundamental insights into the highly dynamic and important process of immune activation. == Project Description == The Computational Immunology group of the Radboud Institute for Molecular Life Sciences aims to model both cellular migration patterns within t cell regions of lymph nodes, and the process of lymph node structure development. Through an international cooperation with labs in Canada and Australia we have access to completely unique data of the three dimensional network structure within mouse lymph nodes. This data allows us to model cell migration in the lymph node at a level that has not been done before, and gives unique insight into how the internal lymph node structure develops. We are looking for someone with a strong interest in computational biology and modelling of cellular behaviour. Initial work on developing data analysis based on deep convolutional neural networks for segmentation of the data and modelling t cell migration using the cellular potts model has been carried out already. You will be involved in taking the data analysis further, coming up with robust ways of quantifying the specific shape and structure of the network, coming up with computational models for development of this network and seeing which model hypothesis best match real data. The work on modelling the shape and development of the network will be combined with work done by us on influence of the shape of the network on antigen presentation efficiency. Our group is also involved with a clinical trial in dendritic cell vaccination, which tries to stimulate the antigen presenting process for cancer patients. This also gives the project immediate practical relevance.

Keywords : Agent Based Model, Computational Biomedicine, Multiscale modelling, Computational Biology,

Track : Computational Biology

Contact 1 : Shabaz Sultan

Valid Date : Dec. 31, 2020

Physics informed neural networks (PINNs) for PDEs and its application on ISR2D for surrogate modeling​

PINNs is a deep neural network framework to solve nonlinear partial differential equations. Such data-driven method can work as an efficient surrogate model for the original PDEs problems if there are sufficient data for training. The goal of the work is to explore PINNs and apply it to construct a surrogate model for our in-stent restenosis submodel, blood flow solver. With the surrogate model, the computational efficiency of the ISR model can be greatly improved and therefore we can implement a semi-intrusive uncertainty quantification analysis to the model.​ ​

Keywords : Machine Learning, Neural Network, Multiscale modelling, Monte Carlo simulation, surrogate modeling, uncertainty quantification,

Track : Others

Contact 1 : Dongwei Ye
Contact 2 : Pavel Zun

Valid Date : Dec. 31, 2020