Abstract:

Immunotherapies have achieved remarkable success in treating hematological cancers, yet the solid tumor microenvironment remains a significant barrier to therapeutic breakthroughs. Machine learning (ML)-driven computational protein design offers a powerful approach to creating novel protein components tailored for specific functions. By combining ML-driven design with synthetic biology and immuno-engineering, we have developed innovative tools to overcome therapeutic barriers within the tumor microenvironment while deepening our understanding of the biological systems involved. In the first half of this seminar, I will introduce three new technologies developed using these approaches: (1) a receptor degradation platform to remodel the tumor microenvironment and enhance therapy, (2) a novel class of receptors that sense soluble tumor markers and localize therapeutic activity, and (3) de novo-designed many-to-many protein networks resembling mammalian signaling complexity. In the second half, I will present a computational protein design pipeline that reconstructs cytokines from the ground up. This approach allows us to design cytokines with customizable receptor clustering geometries, enabling precise exploration of their cell-specific effects and the ability to fine-tune their naturally pleiotropic activities. Furthermore, this approach enabled us to transcend nature’s designs by creating entirely novel cytokines that cells have never encountered before, thus unlocking new therapeutic opportunities to manipulate the tumor microenvironment and expand our understanding of the underlying signaling mechanisms.