FEDKEA: Enzyme function prediction with a large pretrained protein language model and distance-weighted k-nearest neighbor

Recent advancements in sequencing technologies have led to the identification of a vast number of hypothetical proteins, surpassing current experimental capabilities for annotation. Enzymes, crucial for diverse biological functions, have garnered significant attention; however, accurately predicting enzyme EC numbers for proteins with unknown functions remains challenging. Here, we introduce FEDKEA, a novel computational method that integrates ESM-2 and distance-weighted KNN (k-nearest neighbor) to enhance enzyme function annotation. FEDKEA first employs a fine-tuned ESM-2 model with four fully connected layers to distinguish from other proteins. For predicting EC numbers, it adopts a hierarchical approach, utilizing distinct models and training strategies across the four EC number levels. Specifically, the classification of the first EC number level utilizes a fine-tuned ESM-2 model with three fully connected layers, while transfer learning with embeddings from this model supports the second and third-level tasks. The fourth-level classification employs a distance-weighted KNN model. Compared to existing tools such as CLEAN and ECRECer, two state-of-the-art computational methods, FEDKEA demonstrates superior performance. We anticipate that FEDKEA will significantly advance the prediction of enzyme functions for uncharacterized proteins, thereby impacting fields such as genomics, physiology and medicine. FEDKEA is easy to install and currently available at: https://github.com/Stevenleizheng/FEDKEA