Table 3.

Summary of model-generated hypotheses from examining different underlying effects/metrics with the computational model, along with possible experiments and evidence (where available) in the literature in agreement with prediction

Effect/metric examined	Model predicted effect on treatment efficacy	Hypothesis generated	Experimental and/or clinical evidence
T cell recruitment and antigen specificity (Figure 6A)	Increasing the virus- and tumor-antigen specific CD8+ T cells reduces the number of glioblastoma cells	Improvements to OV efficacy are only seen with large populations of virus- and antigen-specific CD8+ T cells	A study in 56 patients with small-cell lung carcinoma (SCLC) detected that a significantly low tumor size was associated with a high number of tumor-infiltrating T cells (Eerola et al., 2000). Staining for CD4+ and CD8+ T cells in 284 gliomas revealed that the number of CD8+ T cells was inversely correlated with tumor grade (Han et al., 2014).
Virus binding rates to stromal cells, $u_s$, and glioblastoma cells, $u_g,$ (Figure 6B–6E)	Increasing the virus binding of cells negatively affects sparse samples and positively affects dense samples	The loss of virus to stroma can significantly hinder OV efficacy	OV-expressed factors such as proteases (e.g. MMP9) have been shown to facilitate stroma degradation, and enhance HSV delivery, distribution, and oncolytic effect in mice bearing human brain tumors (Hong et al., 2010). An EGFR-retargeted HSV with MMP9, KMMP9 demonstrated selective infection and killing and improved viral penetration and survival in a glioblastoma multiforme xenograft model (Sette et al., 2019).
Viral diffusivity, $D_{virus}$, and clearance, ${\lambda }_{virus}$, (Figure 5)	Increased glioblastoma cell death with increases in $D_{virus}$ and decreases in ${\lambda }_{virus}$	Increasing the diffusivity of an OV and/or decreasing the clearance, will significantly improve treatment efficacy	Arming OVs with a transgene that degrades the ECM (e.g. hyaluronidase (Ganesh et al., 2008; Martinez-Quintanilla et al., 2015), decorin (Choi et al., 2010; Zhang et al., 2020) or relaxin ) improves viral spread and enhance viral potency. In glioblastoma, clearance of oHSV has been linked to (natural killer cells) NKs causing diminished OV-efficacy (Alvarez-Breckenridge et al., 2012). Combining TGF-$\text{β}$ with oHSV was shown to inhibit NK recruitment and increased viral efficacy (Han et al., 2015).
Injection location (Figure 7B)	OV administered on the tumor periphery is more effective at reducing the number of glioblastoma tumor cells than OV injected in the center	Intratumoral OV injections are not as effective as injections on the tumor periphery	While there is extensive discussion around the advantages and challenges of intratumoral injections compared with intravenous injections (Geletneky et al., 2017; Zamarin and Pesonen, 2015), experiments that compare administering treatment directly at the tumor periphery versus intratumorally are absent in the literature. Future work could investigate this with a tumor spheroid model.
Proportion of dense to sparse regions in tumor tissue (Figure 7A)	Increased glioblastoma cell death with increased proportion of dense tumor tissue (dense:sparse)	Glioblastoma cell density in the tumor is an indicator of OV efficacy and treatment success	Isolated brain metastatic adenocarcinoma treated with oHSV-1 showed higher HSV-1 infiltration and cleaved caspase-2 in areas of high tumor cell density compared to areas of low tumor cell density (Figures 7C–7F)