Glioblastoma is the most common malignant brain tumor in adults. It is essentially incurable, with a median survival of 14 months from diagnosis. Unlike other difficult forms of cancer, glioblastoma kills patients not by distant metastasis but by rapid local invasion, leading to edema and disruption of vital structures. In the growing tumor, the cancer cells exploit several different routes of invasion. In the same tumor, some cells migrate through the perivascular spaces, whereas others follow white matter tracts, and yet others seed the brain via the cerebrospinal fluid circulation. Invasion along these different routes, which also differs between patients, is very likely mediated by distinct signaling pathways, transcriptional programs, and micro-environmental interactions, each of which may present a new therapeutic opportunity. Understanding these processes and identifying their weak points would open for a new class of therapies. So far, the lack of valid functional models has made it hard to identify and target glioblastoma invasion pathways in a systematic fashion. To solve this, my laboratory has developed a panel of 100 patient-derived glioblastoma xenograft models, which we have characterized at several genomic (DNA sequencing, epigenomics, gene expression, CNA, microRNA) and functional levels (drug Screening and invasion) (Cell Rep 2020 and unpublished work). Our xenograft panel reliably models different routes of invasion, both between and within patient cases, and thereby provides a powerful tool to help us understand invasion in vivo, its genetic regulation, and how it can be blocked pharmacologically. Exploring our resource, we have used single-cell profiling and machine learning (Nat Comm 2020 etc) to identify cell populations and genes that are likely linked to perivascular and perineural invasion. Here, our goal is to is establish individual based in silico simulations of glioblastoma xenograft growth. In our background work, we have formulated a mathematical framework that captures the growth and invasion of individual tumor cells, within the brain anatomy. The anatomy in which the simulated tumor grows is represented by high resolution Magnetic Resonance Imaging data (to describe nerve fiber tracts) and by CLARITY microscopy data (to describe the vasculature). The growth and invasion is simulated by a stochastic algorithm (an adaptation of Gillespies algorithm). Our specific goals with this small compute project are to: 1) set up simulation of brain tumor xenografts at Uppmax 2) evaluate the effects of different simulations parameters on tumor growth 3) use approximate Bayesian computation (ABC) to fit parameters to observational data from individual patient cases. 4) integrate existing anatomical and molecular data sets to support computations