Project Summary A staggering number of potential targets for combating diseases fall into a category called undruggable, targets that are not accessible to two commonly used drug modalites: antibodies and small molecules. These undruggable targets often reside inside the cells and cannot be accessed by antibodies, which are too large to cross the cell membrane. Their flat surfaces deviate from the common deep pockets that are often hit by small molecules. To untap the potential of these targets, we need a therapeutic modality that can target shallow pockets and cross the cell membrane. Peptides are at the size range and composition that can be an ideal choice as this alternative modality. Despite efforts in developing high affinity peptide binders, generating permeable binders has been a long-standing challenge. The few examples of cell permeable functional peptides are often developed through error and trial or via many rounds of modification and testing. The challenge in obtaining permeable binders is mainly due to lack of high throughput methods to screen for permeability, which has not only limited the power of library-based screening for obtaining permeable binders, but has also resulted in a limited amount of experimental data from which one can deduce rules that govern a peptide permeable. Thus, the use of peptides as therapeutics to drug the undruggable has remained largely underdeveloped. This proposal uses an interdisciplinary approach to address this challenge. We generate a network to represent the peptide space in a meaningful manner. We then cluster this network and select for experimental testing a set that is truly representative of the peptide space. The results of our tests, which include both artificial membrane analysis and cellular uptake in bacterial and mammalian cells, can thus be generalized to the entire space. This unprecedented dataset will then be used as a starting point to gain better understanding of peptide permeability and to develop experimental and computational methods for rapid identification of permeable peptide binders that can be used as leads for therapeutic development. We will use this dataset to develop an automated generative algorithm that computationally designs permeable peptides that can target a target interface of interest. This proposal is a leap in the field of peptide therapeutic development. The dataset generated during this work and methods we develop will be the stepping stone for many researchers interested in drug discovery, physics-based models of permeability, and peptide therapeutics.

Project Narrative A large portion of potential drug targets are inaccessible to the most widely used therapeutic modalities: small molecules and antibodies. Peptides are powerful alternatives for drugging these targets; however, their use has been hugely underexplored due to challenges in the way of obtaining cell-permeable peptides that can access intracellular targets. In this proposal, we use a novel data-driven approach that combines state-of-the-art computational techniques, such as peptide design and machine learning, with experimental data generation to develop methods and datasets that will pave the way for the development of next generation cell-permeable peptide therapeutics.