FAD-BERT

FAD-BERT predicts Flavin Adenine Dinucleotide (FAD) binding sites in electron transport chain proteins using pre-trained deep bidirectional transformers and integrated bioinformatics features.

Key Features:

• Deep Bidirectional Transformers: Employs pre-trained BERT models with contextualized word embedding capabilities for protein sequence analysis.
• Integration with Bioinformatics Databases: Incorporates Position-specific Scoring Matrix (PSSM) profiles and the Amino Acid Index database (AAIndex) as input features.
• Predictive Performance: Reports 85.14% accuracy and a Matthew's correlation coefficient (MCC) of 0.39 on an independent dataset, representing an 11% improvement over previous methods.

Scientific Applications:

• Electron Transport Chain Analysis: Identification of FAD-binding sites to inform studies of electron transfer within protein complexes involved in cellular respiration.
• Molecular and Biochemical Research: Supports investigations into protein–cofactor interactions and cellular energy production mechanisms in molecular biology and biochemistry.

Methodology:

Combines pre-trained deep bidirectional transformers (BERT) with contextualized sequence embeddings and integrates PSSM and AAIndex features; evaluated on an independent dataset reporting accuracy 85.14% and MCC 0.39.