Computational modeling of interactions between cell surface proteins with multiple domains in the immunoglobulin fold

About This Grant

Project Summary Cells adapt to their surrounding environments by forming dynamic contact with each other. These contacts are maintained by molecular interactions between receptors and ligands on cell surfaces. The immunoglobulin (Ig) fold, as the largest and most typical class of domains for cell surface recognition, is widely distributed in almost all types of cell surface receptors and their corresponding ligands. Some Ig domains can interact with multiple targets with various binding affinities. The difference in binding specificity of these proteins is a crucial determinant of their biological functions. Moreover, a majority of cell surface proteins contain multiple Ig domains in their extracellular regions. Not all of these domains are directly involved in binding partner recognition. To understand the function of cell surface proteins, it is necessary to determine which specific domains are responsible for binding, how Ig domain binding selectivity is determined, and why multiple extracellular domains need to be evolved. However, it is currently highly challenging to measure the extracellular interactions between cell surface proteins on a systematic level. These interactions are difficult to detect by standard biochemical assays due to the transient nature of their binding kinetics. Computational modeling can reach dimensions that are currently unapproachable in the laboratory. Unfortunately, even the state-of-the-art deep-learning-based methods, such as AlphaFold 3, are not sensitive enough to model the transient interactions among cell surface proteins. Thus, the objective of this proposal is to develop new methods that can be used to predict and simulate the specific interactions between cell surface proteins with Ig domains. We have constructed a non-redundant structural database for Ig domain interactions. Using this database, we will first develop a computational platform that combine protein language model and attention-based deep-learning model to identify potential new interactions between cell surface proteins. Moreover, we will generate a short list of domain pairs between two multi-domain cell surface proteins that are highly likely to mediate their interaction, therefore greatly reducing the complexity of experimental tests. We will further design a two-stage machine-learning strategy to predict the binding constants between specific Ig domains. We will use this method to understand why specific Ig domains can bind to multiple ligands with various affinities. Finally, we will propose a multiscale framework by incorporating protein conformational fluctuations estimated from molecular dynamics simulations into a new mesoscale model. We will use the multiscale framework to explore how multiple Ig domains regulate the intercellular interactions between cell surface proteins. All of our computational predictions will be experimentally validated by our long-term collaborators through an iterative process. Our long-term goal is to elucidate how interactions between cell surface proteins influence and regulate immune signaling and neural plasticity. Ultimately, the computational methods developed in this project are designed to be broadly applicable, enabling the study of protein interactions across diverse domain families beyond our primary systems of interest.