guest ::
| Home > Publications database > On the road to digital twins of tumors |
| Home > Publications database > On the road to digital twins of tumors |
| Dissertation / PhD Thesis | FZJ-2025-02820 |
2025
Universitäts-und Landesblibliothek Düsseldorf
Düsseldorf
This record in other databases:
Please use a persistent id in citations: https://nbn-resolving.org/urn:nbn:de:hbz:061-20250610-092027-1 doi:10.34734/FZJ-2025-02820
Abstract: Many multicellular organisms are composed of tissues. Tissues are groups of similar cells with one or more specific tasks, and it is the combination of multiple tissues that leads to the formation of organs and larger structures within the body. As the complexity of an organism increases, so does the number of different tissues. Tissue generation and differentiation thus plays crucial roles covering the range from life processes from its early stages during embryogenesis to necessary maintenance such as wound healing. Unfortunately, the proliferation machinery of some cells can become defective due to multiple causes. Through this, the tissue growth process may become flawed and uncontrollable. This can lead to the multi-faceted phenomenon known as cancer. Cancer is a disease that can occur in the entire body, and its heterogeneity makes a generally applicable treatment difficult. While understanding the detailed origins of the cell proliferation defects on a genetic and epigenetic level is crucial to stopping the disease before it starts, doctors and patients are often faced with the fact that it has already progressed to a macroscopic stage ((O(mm)) at the time of diagnosis. In this situation, it is necessary to understand the laws governing the growth of tumors, in order to develop effective treatment strategies. Computational models are a useful tool for this for two main reasons. Firstly, because a mechanistic model capable of reproducing experimentally observed behavior underlines our understanding of the underlying mechanisms. Secondly, because a working model would vastly benefit treatment via quick assessments of options without any side effects or harm to the patient. Hence, work is ongoing to develop so-called digital twins of growing tumors. A challenge in this endeavour is the scale-spanning nature of cancer. This means that any model simulating macroscopic tissue and tumor growth must be capable of doing so at single-cell resolution. Such models are computationally demanding, and often require supercomputing infrastructures to employ. Furthermore, care must be taken to parameterize them correctly, and large amounts of experimental data are required for this. During this doctoral project, I have worked with and extended Cells in Silico (CiS), a highly scalable tissue simulation framework previously developed by my group. CiS is capable of simulating biological tissues composed of tens of millions of cells at subcellular resolution, and is therefore a promising candidate for simulating a digital twin. However, before doing so, it must be extended further, and data for its parameterization must be found. To combat the problem of in vivo data scarcity, I have employed a "divide and conquer" approach, in which I aimed to partially parameterize CiS by focusing on smaller in vitro sub systems, for which data exist. During my studies, I first focused on an investigation of the structural environment of tumors, by working with tumor spheroids grown in collagen matrices of varying density. For this, I performed a large number of spheroid growth simulations, in order to reproduce the behavior of the in vitro spheroids. To analyze the agreement between in vitro and in silico spheroids, we developed the overall deviation score (ODS). The ODS, which is a metric for comparing the structure of two spheroids regardless of their origin, provided an objective function for the parameterization of CiS. During this project, we discovered that CiS needs a more realistic description for the extracellular matrix in order to accurately reproduce spheroid behavior. A project to include such a description is ongoing within my group. In the second part of my project, I focused on the nutrient environment of tumors. Here, I incorporated a set of detailed mouse brain vasculature data into CiS, in order to build a more realistic nutrient environment. I then studied the growth behavior of tumors placed in vascular environments of different density and thickness. Within my simulations, I found that vessel density is the main contributor to final tumor volume. Finally, I focused on the advancement of supercomputing infrastructure by participating in the development of a benchmarking pipeline for the JUPITER supercomputer. Overall, my work has improved CiS, and paved the way for bringing it closer to simulating digital twins of tumors.
|
The record appears in these collections: |
PDF
Fulltext