Trans-MoRFs: A Disordered Protein Predictor Based on the Transformer Architecture.

Area of Science:

• Computational Biology
• Structural Bioinformatics
• Protein Science

Background:

• Intrinsically disordered regions (IDRs) are vital for protein function, with molecular recognition features (MoRFs) playing key roles in interactions and regulation.
• Identifying MoRFs is challenging due to their disorder-to-order transitions, and current computational predictors have limitations in accuracy and sequence length adaptability.

Purpose of the Study:

• To introduce Trans-MoRFs, a novel predictor for identifying MoRFs within protein IDRs.
• To leverage the transformer architecture for improved MoRF prediction accuracy and efficiency across diverse protein sequence lengths.

Main Methods:

• Developed Trans-MoRFs, a predictor utilizing the transformer architecture and its self-attention mechanism.
• The model efficiently captures interactions between distant residues in protein sequences, ensuring stability for various sequence lengths.

Main Results:

• Trans-MoRFs achieved a mean area under the curve score of 0.94 on benchmark datasets, surpassing existing MoRF prediction models.
• The predictor demonstrated superior performance across multiple metrics compared to current combined and single MoRF prediction tools.
• The model exhibits high efficiency and stability for both short and long protein sequences.

Conclusions:

• Trans-MoRFs offers excellent accuracy for predicting MoRFs and other functional elements in disordered protein regions.
• The tool aids in understanding protein functions, pinpointing functional segments, and facilitating the discovery of novel drug targets.
• A web server is available for related research and applications.