Developing Modality Support for PROTACs

PROTACs represent a very promising emerging drug modalities. Traditional modalities, small-molecule drugs, for example, or monoclonal antibodies, are limited to certain modes of action, like targeting specific receptors or blocking particular pathways. Due to its ability to target proteins that have been traditionally difficult to target with conventional methods, PROTACs have attracted substantial interest from both academic researchers and pharmaceutical companies, with several PROTACs currently in clinical trials for various indications, including cancer and neurodegenerative disorders. However, the exploration of PROTACs is still relatively in its infancy. With many open research questions, my proposal implements a tutorial, a common feature in DeepChem, to explore how to computational investigate PROTACs. More specifically, I provide a guide towards exploring available PROTAC datasets, developing featurization capabilities, and a multimodal sequence and structure model to infer protein degradation abilities of PROTACs, a clinically relevant biopharmaceutical property. My tutorial will be implemented through a Jupyter notebook which highlights the convenience and low-code capabilities of DeepChem for the scientific community. To demonstrate this, I will implement a few under-the-hood features to process available PROTAC datasets. My tutorial notebook can easily load data and build upon available models in DeepChem to develop a multimodal model for downstream inference. I hope this project provides a strong foundational understanding for researchers and the scientific community to explore computational methods for PROTAC design and optimization, possibly contributing to more effective research and development with downstream clinical benefits. Overall, I believe this project will contribute to the overall mission of DeepChem by democratizing computational life science research for all.