Sampling and scoring: a marriage made in heaven

Area of Science:

• Computational Biology
• Structural Biology
• Biophysics

Background:

• Protein docking algorithms typically involve conformational sampling followed by scoring and refinement.
• Decoupling sampling and scoring is common for developing new methods, often using decoy sets.

Purpose of the Study:

• To investigate the impact of integrating versus decoupling sampling and scoring in protein docking.
• To analyze the role and pitfalls of decoy sets in scoring function development.
• To propose pathways for better alignment of sampling and scoring steps.

Main Methods:

• Review of decoy sets in protein-protein docking.
• Analysis of three case studies on integrated and decoupled docking approaches.
• Investigation of training and decoy set selection effects on scoring functions.
• Examination of algorithms for integrating sampling and scoring.

Main Results:

• Complete decoupling of sampling and scoring is suboptimal for realistic protein docking problems.
• Integrated sampling and scoring generally outperform methods that only score pre-generated structures.
• The choice of training and decoy sets critically impacts scoring function performance.

Conclusions:

• Integrating sampling and scoring steps leads to substantial improvements in protein docking accuracy.
• Careful consideration of decoy sets is crucial for effective scoring function development.
• The findings are applicable to other conformational search problems, including protein structure prediction.