Toward model-driven development of viral therapies

Abstract

Oncolytic virotherapy is a promising cancer treatment that uses viruses to selectively kill cancer cells. Historically, viral design has been largely qualitative, disrupting genes that help viruses evade the immune system and hijack cellular machinery and inserting pro-inflammatory genes. Here, I present a quantitative framework for tuning gene expression in common oncolytic viruses. First, I introduce a model that predicts viral replicative fitness as a function of transcription rates. I then propose a transcription mechanism in these viruses that enables more precise control to further optimize desirable phenotypes. Generating viruses from plasmids typically relies on many plasmids (3-8). Ensuring that every plasmid enters the same cell and expresses its gene is inherently inefficient. I revisit the assumption that multiple small plasmids are more efficient than a few larger ones. I show that plasmid entry rates are effectively size-independent across ~5 to 16 kb, that lipoplexes co-deliver only limited multi-plasmid cargo, and that consolidating helper genes onto single, larger plasmids increases both the probability of co-expression and the correlation between expressed genes. Finally, using a spatial agent-based tumor model, I demonstrate that maximizing infectivity in cancer cells alone is insufficient for tumor clearance in stroma-rich tissues. Granting the virus calibrated ability to infect stromal cells markedly improves penetration and overall control, revealing a therapeutic window for partial stromal targeting. Together, these results shift oncolytic virus design from trial-and-error to principled optimization. By mapping transcription to fitness, streamlining the assembly of these viruses, and accounting for tumor microenvironment barriers, this work provides design rules for future oncolytic virotherapies.