Streamlining directed evolution of oncolytic adenoviruses by construction of enhanced viral libraries using adaptive in silico toolsets

• Applicant: Dr. Julian Fischer
• Subject Area: Virology
• Term: since 2026
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 580244154

Human adenoviruses (HAdV) are widely used as viral vectors. In clinical applications, they are extremely popular as vectors for vaccinations or oncolytic therapy. Despite the widespread use of this platform, the application of oncolytic HAdV is severely limited. Currently used recombinant adenoviruses (rAd) mostly rely on naturally occurring tropisms to infect their target cells. This leads to limited treatment efficiency, as tumor cells are not specifically infected. Additionally, systemic applicability is reduced since healthy tissue can also be infected. To address this issue, Ad5NULL, an rAd based on HAdV-C5 that prevents unspecific binding to cellular receptors, was developed. Based on this virus, several novel therapies have already been developed and are currently being tested in clinical trials. However, identifying new cellular target receptors is extremely time-consuming and inefficient. With our new technology for generating genetic libraries, we have already demonstrated that we can achieve the necessary diversity of viral particles to enable in vitro directed evolution. Now, we aim to apply this method using Ad5NULL to develop novel, targeted rAd vectors without unwanted off-target effects. Due to the immense diversity of viral particles required for this approach, we will develop a machine-learning-based algorithm that will use experimentally obtained data to identify correlations between vector efficiency and biochemical characteristics. Given natural biological diversity, it is highly unlikely that this algorithm will be able to predict specific interactions. Instead, we aim to use it to increase the probability of randomly occurring viable variants. Certain characteristics will negatively impact the vector performance of emerging viral particles, which we will filter out in silico. This will reduce the required diversity within the library and improve the ratio of potentially viable variants. Replication within relevant cancer cell lines is to be used to assess vector performance. Instead of focusing solely on receptor binding (as in phage display library applications), this will give us an aggregate factor for binding, infection, transduction, and host cell response to successfully identify translationally relevant vector variants. Through iterative repetitions of this process, the precision of the model will continuously improve. Thus, this project will not only develop novel vector candidates but also provide fundamental tools for vector development and other key applications.

• DFG Programme: Fellowship
• International Connection: United Kingdom
• Host: Professor Alan Parker, Ph.D.