The MULTICOM Protein Structure Prediction Server Empowered by Deep Learning and Contact Distance Prediction

Area of Science:

• Computational Biology
• Structural Bioinformatics
• Deep Learning Applications

Background:

• Protein structure prediction is crucial for understanding biological function.
• Advances in deep learning and coevolutionary analysis have improved protein structure prediction accuracy.
• The CASP13 experiment highlighted progress in both template-based and template-free methods.

Purpose of the Study:

• To provide a practical guide to the latest MULTICOM protein structure prediction system.
• To detail the functionalities of the MULTICOM system, including 1D, 2D, and 3D structure prediction.
• To present the MULTICOM web server and its accessibility for users.

Main Methods:

• Integration of deep learning for contact distance prediction.
• Utilizing residue-residue coevolutionary analysis.
• Employing template-based and template-free (ab initio) modeling approaches.
• Incorporating domain boundary prediction and protein quality assessment.

Main Results:

• The MULTICOM system predicts 1D structural features, 2D interresidue contacts, and 3D structures.
• Domain boundary prediction and template-based modeling are key functionalities.
• Contact distance-driven ab initio modeling and deep learning-enhanced quality assessment are included.
• The system was rigorously tested in the 2018 CASP13 experiment.

Conclusions:

• The MULTICOM system offers a comprehensive suite of tools for protein structure prediction.
• User-friendly web interfaces and accessible source code facilitate the use of MULTICOM.
• The system represents a significant advancement in computational protein structure prediction.