BERT-Kgly: A Bidirectional Encoder Representations From Transformers (BERT)-Based Model for Predicting Lysine Glycation Site for Homo sapiens

• DOI: 10.3389/fbinf.2022.834153
• Publication Date: 2022-02-18T09:12:06Z
• Authors: Yinbo Liu; Yufeng Liu; Gang-Ao Wang; Yinchu Cheng; Shoudong Bi; Xiaolei Zhu
• Keywords: UniProt

As one of the most important posttranslational modifications (PTMs), protein lysine glycation changes characteristics proteins and leads to dysfunction proteins, which may cause diseases. Accurately detecting sites is great benefit for understanding biological function potential mechanism in treatment However, experimental methods are expensive time-consuming site identification. Instead, computational methods, with their higher efficiency lower cost, could be an supplement methods. In this study, we proposed a novel predictor, BERT-Kgly, prediction, was developed by extracting embedding features segments from pretrained Bidirectional Encoder Representations Transformers (BERT) models. Three BERT models were explored get embeddings optimal representability, three downstream deep networks employed build our Our results showed that model based on extracted 556,603 sequences UniProt outperforms other addition, independent test set used evaluate compare existing indicated superior