Chemical-SA-BiLSTM

Chemical-SA-BiLSTM predicts grain protein functions by combining chemical properties with amino acid sequence information to annotate proteins from soybean, maize, indica rice, and japonica rice using genome sequencing data.

Key Features:

• Integration of chemical properties and amino acid sequences: The model fuses protein chemical properties with amino acid sequences as input features.
• Self-Attention Mechanism: A self-attention module weights positions within input sequences to emphasize relevant residues.
• Bidirectional Long Short-Term Memory (BiLSTM): BiLSTM layers capture sequence dependencies in both forward and backward directions.
• Benchmarking and performance: Reported experimental results indicate performance superior to classical neural network algorithms for grain protein function prediction.
• Target species: Applied to proteins from soybean, maize, indica rice, and japonica rice.

Scientific Applications:

• Agricultural genomics and proteomics: Functional annotation of grain proteins to support genomics and proteomics studies.
• Crop improvement and breeding: Identification of protein functions relevant to yield, disease resistance, and stress tolerance for breeding programs.
• Nutritional and molecular analysis: Prediction of protein functions to inform nutritional analysis and molecular studies of plant biology.

Methodology:

The method fuses chemical properties with amino acid sequences within a neural network architecture that employs a self-attention mechanism and BiLSTM layers and evaluates performance against classical neural network algorithms.