The energy computation paradox and ab initio protein folding

Area of Science:

• Computational biochemistry
• Structural biology
• Biophysics

Background:

• Predicting three-dimensional protein structure from amino acid sequence is a fundamental challenge.
• Physics-based scoring functions are widely used, assuming they can distinguish native from non-native protein folds.
• Conformational sampling is often considered the primary bottleneck in protein folding simulations.

Purpose of the Study:

• To investigate the impact of protein size on the accuracy of physics-based scoring functions.
• To analyze the nature and magnitude of errors in computed protein energies.
• To determine the feasibility of accurately predicting the structures of large proteins using current methods.

Main Methods:

• Analysis of error probability density functions for physics-based scores.
• Comparison of computed energies with accurate reference energies.
• Evaluation of energy landscape distortions caused by computational errors.

Main Results:

• Computed energies from physics-based scoring functions contain significant systematic and random errors.
• These errors increase with protein size, becoming a major problem for large proteins.
• Energy landscape distortions due to errors significantly reduce the chances of finding the correct free energy minima.

Conclusions:

• Current physics-based scoring functions have limited success in predicting the structures of large proteins due to inherent errors.
• Understanding and quantifying these errors is crucial for improving energy computation accuracy.
• Post-hoc error reduction strategies can enhance the performance of scoring functions for better fold discrimination.