"Supercomputing + Intelligence" Dual Drive Accelerates PROTAC Drug Development - Sun Yat-sen University News Website

**Abstract:** Researchers from Sun Yat-sen University, led by Professor Yang Yue Dong from the National Supercomputing Center in Guangzhou, in collaboration with the research team from Star PharmaTech, have developed an innovative computational framework to accelerate the design and optimization of PROTACs (Proteolysis Targeting Chimeras). PROTACs, a promising therapeutic strategy, leverage the ubiquitin-proteasome system to degrade disease-causing proteins, which are often considered "undruggable" targets due to their challenging characteristics. However, the complex trident structure of PROTACs, characterized by a higher molecular weight compared to traditional small molecules, poses significant challenges for their optimization, including issues with solubility, permeability, bioavailability, and the unpredictable Hook effect. To address these challenges, the team introduced PROTAC-RL, a rational design algorithm based on deep generative models. This algorithm takes a pair of E3 ligands and warheads as inputs and outputs a designed linker in a reinforcement learning (RL) environment, generating PROTAC molecules with specific desired properties. For the proof-of-concept, the researchers selected BRD4, a well-known target protein, and generated over 5,000 virtual PROTACs. These virtual molecules were further analyzed using machine learning scoring and molecular dynamics simulations, facilitated by the extensive computational resources of the "Tianhe-2" supercomputer and the biomedical application platform at the National Supercomputing Center in Guangzhou. From the initial pool, six PROTACs were chosen for synthesis and biological testing based on their synthetic accessibility. Among these, three demonstrated significant inhibitory activity against BRD4, and one of these compounds exhibited high anti-proliferative effects in tumor cell lines and favorable pharmacokinetic properties in mice. The entire process, from initial design to experimental validation, was completed in just 49 days, highlighting the efficiency and potential of integrating supercomputing, artificial intelligence, and molecular dynamics in the rational design and optimization of PROTACs. This study, published in the international journal *Nature Machine Intelligence*, marks a significant advancement in the development of PROTACs, offering a new approach to overcoming the limitations of traditional empirical methods. The automated computational framework not only accelerates the discovery of novel lead compounds but also enhances the likelihood of successful clinical translation by optimizing key pharmacological properties. The results underscore the importance of interdisciplinary collaboration and the application of cutting-edge computational tools in advancing the field of biopharmaceutical research. **Key Elements:** - **PROTACs (Proteolysis Targeting Chimers):** A novel therapeutic strategy that targets and degrades disease-causing proteins, particularly those that are traditionally challenging to drug. - **National Supercomputing Center in Guangzhou:** Provided the computational resources, including the "Tianhe-2" supercomputer and a biomedical application platform, crucial for the rapid and efficient design and optimization process. - **PROTAC-RL Algorithm:** A deep generative model driven by reinforcement learning, designed to generate optimal linkers for PROTAC molecules. - **BRD4 Target Protein:** Used as a case study to validate the effectiveness of the PROTAC-RL approach. - **Molecular Dynamics Simulations and Machine Learning:** Techniques employed to assess and refine the properties of the generated PROTACs. - **Experimental Validation:** Six synthesized PROTACs were tested, with three showing significant inhibitory activity and one demonstrating favorable pharmacokinetic properties in mice. - **Publication:** The research findings were published in *Nature Machine Intelligence*, emphasizing the potential of the computational framework in biopharmaceutical development. **Significance:** The development of the PROTAC-RL algorithm and its successful application in generating novel lead compounds with high degradation activity and improved pharmacokinetic properties represent a significant breakthrough in the field of PROTAC research. By leveraging the power of supercomputing and artificial intelligence, the team has demonstrated a more efficient and rational approach to drug design, which could potentially lead to faster clinical translation and the discovery of new treatments for diseases that are currently difficult to target. This interdisciplinary effort highlights the importance of combining computational and experimental approaches to overcome the inherent challenges in PROTAC design and optimization.