DMFpred: Predicting protein disorder molecular functions based on protein cubic language model.

Area of Science:

• Biochemistry and Molecular Biology
• Computational Biology
• Genomics and Proteomics

Background:

• Intrinsically disordered proteins and regions (IDPs/IDRs) are vital for cellular processes and implicated in human diseases.
• Current experimental methods for identifying IDP functions are costly and time-consuming, limiting large-scale analysis.
• Existing computational tools primarily focus on predicting only one type of IDP function, leaving a gap in comprehensive prediction.

Purpose of the Study:

• To develop a computational method for predicting five essential molecular functions of intrinsically disordered proteins: assembler, scavenger, effector, display site, and chaperone.
• To address the limitations of experimental techniques and existing predictive models for disordered protein functions.
• To facilitate drug target discovery and rational drug design by providing accurate functional predictions for IDPs.

Main Methods:

• Development of DMFpred, a novel predictor utilizing a Protein Cubic Language Model (PCLM).
• PCLM integrates sequence, structural, and functional features using three protein language models and attention-based alignment.
• The model was pre-trained on extensive IDR sequences and fine-tuned using experimentally annotated functional sequences.

Main Results:

• DMFpred demonstrated high-quality predictive performance across five categories of disordered protein functions.
• The tool accurately predicts both single and multiple functional roles of residues within IDPs.
• Evaluation confirmed the effectiveness of DMFpred in expanding the coverage of disordered protein function prediction.

Conclusions:

• DMFpred offers a robust and scalable computational solution for predicting diverse intrinsically disordered protein functions.
• The tool aids in understanding the roles of IDPs in health and disease, supporting therapeutic strategies.
• DMFpred is accessible via a web server, promoting its use in biological research and drug development.