Building the new MatriNET application for the study of extracellular matrix network dynamics

Abstract

The extracellular matrix (ECM) is a vital three-dimensional structure which provides structural and mechanical support to neighbouring cells, functions as a reservoir for growth-factors and bioactive molecules. As intracellular signalling pathways are strongly influenced by stimuli produced by the ECM, structural and mechanical alterations of the ECM have been associated with many diseases, such as cancer, fibrosis, and osteoarthritis. The comprehensive characterization of the ECM has been limited by the biochemical properties of the ECM. However, given the important role of the ECM in health and disease, there is a compelling reason to study the ECM more carefully. Over the past few years, computational approaches have been developed to overcome the limitations of traditional methods and thus accelerate discoveries related to the ECM. One of these approaches is MatriNET, a specialized tool that considers the network nature of the ECM. This study provides an updated version of MatriNET allowing the generation of real single-cell networks and analysing gene activities in sample datasets. Additionally, the study demonstrates the practical use of the updated application as well as analyzes the results obtained from the application by visualizing them. To achieve this, the application was developed with the R programming language, with some snippets of JavaScript. The development was carried out using the Shiny R package, along with other R packages to incorporate additional functionalities. To identify patterns and clusters from the results obtained from the application, a UMAP visualization was plotted. The results show that similar cell types tend to cluster together, with non-cancer cells showing close similarities within each type. However, some non-cancer cell types also separated into subgroups, indicating heterogeneity within the non-cancer cell populations potentially driven by differences in gene expression or cellular properties. Cancer cells similarly segregate into subgroups. The largest cluster is distinctly separated from the non-cancer cells, implying that most of the cancer cells have their own gene activity patterns. However, smaller subgroups of cancer cells are located more closely to some of the non-cancer cell clusters, indicating possible gene expression similarities or interactions related to their roles in the tumour microenvironment. These findings demonstrate that the updated MatriNET application successfully differentiates cell types, revealing structural relationships and heterogeneity within both cancer and non-cancer cells in the example dataset. Overall, this study contributes to exploring the ECM as it provides insight into structural changes in the ECM network.