Integration of transcriptomics data into agent-based models of solid tumor metastasis

Most of the recent progress in our understanding of cancer relies in the systematic profiling of patient samples with high throughput techniques like transcriptomics. This approach has helped in finding gene signatures and networks underlying cancer aggressiveness and therapy resistance. However, -omics data alone is not sufficient to generate insights into the spatiotemporal aspects of tumor progression. Here, multi-level computational models are promising approaches, which would benefit from the possibility to integrate in their characterization the data and knowledge generated by the high throughput profiling of patient samples. We present a computational workflow to integrate transcriptomics data from tumor patients into hybrid, multi-scale models of cancer. In the method, we employ transcriptomics analysis to select key differentially regulated pathways in therapy responders and non-responders and link them to agent-based model parameters. We next utilize global and local sensitivity together with systematic model simulations to assess the relevance of variations in the selected parameters in triggering cancer progression and therapy resistance. We illustrate the methodology with a de novo generated agent-based model accounting for the interplay between tumor and immune cells in melanoma micrometastasis. Application of the workflow identifies three different scenarios of therapy resistance.