Search for: All records

Recently, targeted degradation has emerged as a powerful therapeutic modality. Relying on “event-driven” pharmacology, proteolysis targeting chimeras (PROTACs) can degrade targets and are superior to conventional inhibitors against undruggable proteins. Unfortunately, PROTAC discovery is limited by warhead scarcity and laborious optimization campaigns. To address these shortcomings, analogous protein-based heterobifunctional degraders, known as bioPROTACs, have been developed. Compared to small-molecule PROTACs, bioPROTACs have higher success rates and are subject to fewer design constraints. However, the membrane impermeability of proteins severely restricts bioPROTAC deployment as a generalized therapeutic modality. Here, we present an engineered bioPROTAC template able to complex with cationic and ionizable lipids via electrostatic interactions for cytosolic delivery. When delivered by biocompatible lipid nanoparticles, these modified bioPROTACs can rapidly degrade intracellular proteins, exhibiting near-complete elimination (up to 95% clearance) of targets within hours of treatment. Our bioPROTAC format can degrade proteins localized to various subcellular compartments including the mitochondria, nucleus, cytosol, and membrane. Moreover, substrate specificity can be easily reprogrammed, allowing modular design and targeting of clinically-relevant proteins such as Ras, Jnk, and Erk. In summary, this work introduces an inexpensive, flexible, and scalable platform for efficient intracellular degradation of proteins that may elude chemical inhibition.

When aiming to make predictions over targets in the pharmacological setting, a data‐focused approach aims to learn models based on a collection of labeled examples. Unfortunately, data sharing is not always possible, and this can result in many different models trained on disparate populations, leading to the natural question of how best to use and combine them when making a new prediction. Previous work has focused on global model selection or ensembling, with the result of a single final model across the feature space. Machine‐learning models perform notoriously poorly on data outside their training domain, however, due to a problem known as covariate shift, and so we argue that when ensembling models the weightings for individual instances must reflect their respective domains—in other words, models that are more likely to have seen information on that instance should have more attention paid to them. We introduce a method for such an instance‐wise ensembling of models called Synthetic Model Combination (SMC), including a novel representation learning step for handling sparse high‐dimensional domains. We demonstrate the use of SMC on an example with dosing predictions for vancomycin, although emphasize the applicability of the method to any scenario involving the use of multiple models.