caRBP-Pred: Leveraging Protein Language Models for the Prediction of Chromatin-Associated RNA-Binding Proteins

RNA-binding proteins (RBPs) play pivotal roles in cellular processes ranging from RNA metabolism to 3-dimensional genome organization. A distinct subset, chromatin-associated RBPs (caRBPs), binds directly to chromatin to function as transcriptional regulators. However, experimental identification of caRBPs using techniques such as chromatin immunoprecipitation sequencing and mass spectrometry is labor-intensive and costly. Existing computational tools for DNA-binding protein and RBP prediction often rely on outdated Gene Ontology annotations and fail to capture the unique characteristics of chromatin association. Here, we introduce caRBP-Pred, a deep learning framework that integrates a pre-trained protein language model with convolutional neural networks and bidirectional long short-term memory networks. By leveraging full-length protein sequences and evolutionary embeddings from ProtT5-XL, caRBP-Pred significantly outperforms existing DNA- and RNA-binding protein predictors. Application of our model to the mouse proteome identified 41 high-confidence caRBP candidates. Multidimensional validation using the COMPARTMENTS database and InterProScan confirmed that a proportion of these candidates possess experimentally verified chromatin-binding domains or high-confidence nuclear localization. Collectively, caRBP-Pred is the first computational tool specifically designed for caRBP prediction, offering a valuable resource for investigating the regulatory roles of caRBPs in chromatin-related function.