Impact of massively parallel computation on protein structure determination

Area of Science:

• Protein chemistry
• Computational biology
• Biophysics

Background:

• A long-standing principle posits that biologically active proteins exist at thermodynamic equilibrium, adopting their minimum free energy structure.
• While often true and guiding structure determination, recent evidence suggests this may not apply to all proteins.
• This principle is crucial for computational and experimental (diffraction, resonance) structure determination methods.

Purpose of the Study:

• To review the challenges in general protein structure prediction from sequence.
• To discuss the impact of increasing computer power on addressing these challenges.
• To explore new approaches for solvation and parallel algorithm design in protein modeling.

Main Methods:

• Review of existing methodologies and theoretical frameworks for protein structure prediction.
• Discussion of computational limitations, including potential functions and phase space sampling.
• Exploration of advanced computational techniques, such as parallel algorithms and improved solvation models.

Main Results:

• Key difficulties in protein structure prediction include inadequate potential functions (especially regarding solvent effects) and inefficient sampling of conformational space.
• Ensuring identification of the global minimum free energy state remains a significant computational hurdle.
• Experimental data can mitigate these issues, but challenges intensify with fewer empirical constraints.

Conclusions:

• Understanding the theoretical underpinnings of effective theories is essential for accurate protein structure prediction.
• Numerical methods must effectively integrate the interplay of various temporal and spatial scales.
• Increased computational power, coupled with novel approaches to solvation and parallel computing, offers promising avenues for advancing protein structure prediction.