Selection of representative structures from large biomolecular ensembles

Area of Science:

• Computational Biology
• Structural Biology
• Bioinformatics

Background:

• Protein structure determination is challenging despite experimental advances.
• Computational simulations complement experimental data, especially for sparse or low-resolution information.
• Selecting representative structures from large computational ensembles remains a key challenge.

Purpose of the Study:

• To introduce a novel method for selecting representative protein structures from computationally generated ensembles.
• To utilize co-evolutionary contact pairs as distance restraints within a physical force field to guide simulations.
• To evaluate the effectiveness of ensemble selection algorithms in identifying native-like protein folds.

Main Methods:

• Employed replica-exchange molecular dynamics (REMD) simulations across a wide temperature range for five mid-sized proteins.
• Integrated bias derived from co-evolutionary contact pairs, predicted by a deep residual neural network, to guide simulations.
• Developed and applied four robust ensemble-selection algorithms to extract representative structural models.

Main Results:

• Contact-guided REMD simulations successfully generated ensembles biased towards native-like conformations.
• The study demonstrated the precision of this method for mid-sized proteins.
• Ensemble selection algorithms proved capable of extracting representative structural models with high certainty, even in blind scenarios.

Conclusions:

• The proposed computational method effectively guides protein structure simulations using co-evolutionary data.
• Robust ensemble-selection algorithms are crucial for accurately identifying native-like protein folds from simulation data.
• This approach enhances the interpretation of sparse or low-resolution experimental data in structural biology.