EnAcrPred: A robust ensemble machine learning framework for identifying anti-CRISPR proteins

The identification of anti-CRISPR proteins (Acrs) is crucial for understanding the regulation of CRISPR-Cas systems and their application in gene editing. However, current experimental methods face challenges, particularly in detecting Acrs with low similarity to known protein sequences. To address these challenges, we propose EnAcrPred, an advanced prediction framework based on ensemble learning. The model combines features such as sequence composition, order correlation, and inferred structure and utilizes a stacking ensemble architecture to integrate multiple base models, which enhances both the accuracy and generalization ability of the predictions. Experimental results demonstrate that EnAcrPred achieves superior performance over existing methods across multiple evaluation metrics, further confirming its robustness. Additionally, SHapley Additive exPlanations (SHAP) value analysis identifies the key features influencing Acrs recognition. To facilitate broad adoption in practice, we developed an online platform where users can quickly obtain Acrs predictions by entering a protein sequence. EnAcrPred offers an effective solution for Acrs identification, contributing to the advancement of gene editing research and safety. The platform is accessible via the link at https://ycclab.cuhk.edu.cn/EnAcrPred/.