Precise control over CRISPR systems is critical for both biological discovery and therapeutic translation, yet natural anti-CRISPR proteins are sparse, particularly for RNA-targeting Type VI Cas13 enzymes[1-3]. To address this gap, we designed potent protein inhibitors of Cas13, demonstrating that artificial intelligence–guided protein design can replace discovery-based approaches for generating functional regulators of complex biomolecular machines[4].
Using structure-guided diffusion models combined with sequence optimization, we designed small, stable protein inhibitors targeting the HEPN nuclease site of Leptotrichia buccalis Cas13a, a system for which no natural inhibitors are known. From a focused experimental screen, multiple designs exhibited strong inhibitory activity, with lead candidates achieving low-nanomolar potency in vitro. Biochemical and structural analyses confirmed competitive inhibition of the catalytic site without disruption of guide RNA binding. These AI-designed anti-CRISPRs function in bacterial phage-defence assays and in mammalian cells, restoring gene expression while maintaining favourable biophysical properties.
Beyond the immediate biological outcome, this work served as a foundational demonstration for the establishment of the AI-Protein Design Program (AIPDP) at Monash University. Building on this initial success, AIPDP has evolved into a scalable, modular pipeline integrating binder design, experimental screening, and mechanistic validation, supporting diversification across protein architectures, binding modes, and applications[5].