Mehmet Izmirli* and Pinar Pir
Gebze Technical University
mehmettizmirli [at] gmail.com
Abstract
Agent-based modelling (ABM) is a computational approach in which system-level dynamics are simulated by modelling each component as individuals, defined by a relatively simple set of rules. The ABM method is well-suited for modelling heterogeneous tumour microenvironments (TME), in which cells interact with their environment and with other cells. The progression of the disease is significantly affected by the features of the resident tissue, such as architecture and physiological conditions. This is especially true for metabolically complex organs like the liver. Most cancer ABMs, however, are built based on simplifications that limit biological realism.
In this study, an omics data-driven workflow was established to guide ABM development, with a focus on hepatocellular carcinoma (HCC). First, a single-cell RNA sequencing (scRNAseq) dataset of HCC, including tumour and healthy liver samples, was processed and integrated to remove batch effects between samples. Using the integrated data, major cell types, like malignant cells, were annotated. Then, spatial deconvolution of a spatial transcriptomics (ST) dataset was performed using the scRNAseq dataset as a reference, to predict the cell-type composition and spatial distribution over a tissue slice. Lastly, these predictions, along with the physical measurements of the tissue slice, were used to calculate the spatial coordinates of individual cells.
The resulting coordinates were then used to initialise the agent positions in the ABM, while the behaviour of individual agents was modelled based on the literature and transcriptomic features obtained from scRNAseq data. Finally, to validate the credibility of the tumour growth patterns, preliminary simulations were run, and a qualitative comparison was made with observed tumour regions, including the tumour core and the growing edge. This established framework allows characterisation of emergent tumour behaviours during malignant growth and spread in the context of tissue-specific spatial heterogeneity introduced with spatial data integration.
Thus, an agent-based model of hepatocellular carcinoma, informed by integrated scRNAseq and ST datasets, was developed, promising more realistic and data-driven simulations of cancer dynamics and metastatic potential.
Keywords: Agent-Based Modelling, scRNAseq, Spatial Transcriptomics
Acknowledgement: This project is generously funded by TÜBİTAK UIDB 1071 ( ERA-NET TRANSCAN-3 , 122N905)