HyperAttention and Linformer-Based β-catenin Sequence Prediction For Bone Formation.

Area of Science:

• Biochemistry
• Bioinformatics
• Molecular Biology

Background:

• Beta (β)-catenin is crucial for bone development and homeostasis.
• Dysregulation of β-catenin is linked to various bone disorders.
• Peptide therapeutics offer targeted treatment with potentially fewer side effects.

Purpose of the Study:

• To develop a HyperAttention and informatics-based model for predicting β-catenin peptide sequences.
• To enhance understanding of β-catenin's role in bone formation through sequence prediction.
• To explore advanced computational methods for peptide drug discovery in bone diseases.

Main Methods:

• Downloaded and quality-checked β-catenin protein sequences from UniProt.
• Utilized DeepBio for predictive analysis and data preprocessing (duplicate, outlier, missing value checks).
• Applied HyperAttention and Linformer models to encoded peptide sequences after splitting data into 80% training and 20% testing sets.

Main Results:

• The HyperAttention model achieved 87% prediction accuracy.
• The Linformer model demonstrated 89% prediction accuracy.
• Both models exhibited high sensitivity and specificity, with Linformer showing superior negative instance identification (91%) and slightly better overall sensitivity.

Conclusions:

• HyperAttention and Linformer models are effective for predicting β-catenin peptide sequences with high accuracy.
• These models show promise for developing targeted peptide therapeutics for bone disorders.
• Further research is needed to optimize model performance and balance positive/negative instance prediction for clinical application.