GL-BLSTM: a novel structure of bidirectional long-short term memory for disulfide bonding state prediction

TL;DR

This paper introduces GL-BLSTM, a novel nested bidirectional LSTM neural network architecture that significantly improves the accuracy of disulfide bonding state prediction in proteins by capturing both local and global sequence features.

Contribution

The paper presents a new nested BLSTM structure, GL-BLSTM, that effectively integrates local and global features for disulfide bond prediction, outperforming existing methods.

Findings

Achieved 90.26% residue-level accuracy

Reached 83.66% protein-level accuracy

Outperformed previous state-of-the-art methods

Abstract

Background: Disulfide bonds are crucial to protein structural formation. Developing an effective method topredict disulfide bonding formation is important for protein structural modeling and functional study. Mostcurrent methods still have shortcomings, including low accuracy and strict requirements for the selection ofdiscriminative features. Results: In this study, we introduced a nested structure of Bidirectional Long-short Term Memory(BLSTM)neural network called Global-Local-BLSTM (GL-BLSTM) for disulfide bonding state prediction. Based on thepatterns of disulfide bond formation, a BLSTM network called Local-BLSTM is used to extract context-basedfeatures around every Cys residue. Another BLSTM network called Global-BLSTM is introduced aboveLocal-BLSTM layer to integrate context-based features of all Cys residues in the same protein chain, therebyinvolving inter-residue relationships in the training process. According to our experimental results,GL-BLSTMnetwork performs much better than other methods, GL-BLSTM reached 90.26% accuracy at residue-level and 83.66% at protein-level. This model has reached the state of the art in this field. Conclusion: GL-BLSTMs special structure and mechanisms are beneficial to disulfide bonding state prediction.By applying bidirectional LSTM, it can extract context based features by processing protein sequences, therebyobtain more discriminative information than traditional machine learning methods. Whats more, GL-BLSTMsspecial two-layer structure enables it to extract both local and global features, in which global features areplaying important roles in improving prediction accuracy, especially at protein-level. Keywords: Disulfide bonds; Prediction; Deep learning; Bidirectional long short term memory