Prediction of Structures and Interactions from Genome Information.

Area of Science:

• Biophysics
• Computational Biology
• Structural Biology

Background:

• Predicting three-dimensional residue-residue contacts from protein sequences using evolutionary information has been explored since the early 1990s.
• Historically, contact prediction accuracy was low (<20% true positives) in major evaluations like CASP before CASP11.
• Recent advancements have dramatically improved contact prediction, enabling accurate three-dimensional protein model generation.

Purpose of the Study:

• To review statistical methods for extracting causative correlations from evolutionary data.
• To discuss various approaches for describing protein structure, complexes, and flexibility using predicted contacts.

Main Methods:

• Disentangling direct from indirect correlations in amino acid covariations/cosubstitutions.
• Employing statistical methods to analyze evolutionary information.
• Developing approaches to model protein structure, complexes, and flexibility.

Main Results:

• Significant improvements in contact prediction accuracy have been achieved.
• Accurate three-dimensional protein models can now be generated from predicted contacts.
• Distinguishing direct from indirect evolutionary correlations is key to this improvement.

Conclusions:

• Advanced statistical methods have revolutionized protein contact prediction.
• Predicted contacts are now a powerful tool for determining protein structure, complexes, and flexibility.
• This progress opens new avenues for understanding protein function and design.