DESTINI: A deep-learning approach to contact-driven protein structure prediction.

Area of Science:

• Computational biology
• Structural biology
• Bioinformatics

Background:

• Protein structure prediction from amino acid sequence is a fundamental challenge.
• Existing template-based methods face limitations with difficult protein targets.

Purpose of the Study:

• To introduce DESTINI, a novel computational approach for protein tertiary structure prediction.
• To demonstrate DESTINI's significantly improved predictive ability on a large scale.

Main Methods:

• Combining a deep-learning algorithm for contact prediction with template-based modeling.
• Applying DESTINI to predict tertiary structures for over 1,200 single-domain proteins.

Main Results:

• DESTINI successfully predicts tertiary structures for four times more "hard" targets than previous methods.
• Improved model quality for both "easy" and "hard" protein targets.
• Deep learning predicts native-like contact patterns, outperforming co-evolutionary analysis.

Conclusions:

• DESTINI represents a significant advancement in protein structure prediction.
• The integration of deep learning-based contact prediction enhances template modeling.
• This approach offers a promising strategy towards solving the protein structure prediction problem.