Analysis of basic clustering algorithms for numerical estimation of statistical averages in biomolecules

Area of Science:

• Statistical mechanics
• Computational physics
• Biomolecular simulations

Background:

• Calculating equilibrium properties in statistical mechanics involves averaging over numerous microstates, which is computationally intractable.
• Clustering algorithms offer a method to approximate these calculations by dividing systems into smaller, manageable clusters.
• While used in biomolecular computations, clustering algorithms are relatively unexplored in this specific context.

Purpose of the Study:

• To theoretically analyze the error and computational complexity of basic clustering algorithms in biomolecular electrostatics.
• To derive a tight and computationally inexpensive error bound for equilibrium states calculated using these algorithms.
• To investigate the relationship between the derived error bound and root mean square error for practical applications.

Main Methods:

• Theoretical analysis of error and computational complexity for two basic clustering algorithms.
• Derivation of a computationally inexpensive error bound for equilibrium states.
• Empirical analysis to establish the relationship between error bound and root mean square error.

Main Results:

• A tight, computationally inexpensive error bound was derived for equilibrium states computed via clustering algorithms.
• A strong empirical relationship was found between the error bound and root mean square error.
• The derived error bound can serve as a predictive metric for algorithm accuracy.

Conclusions:

• Clustering algorithms, with accurate error bounds, can be sufficiently precise for practical biomolecular applications.
• The derived error bound offers a computationally efficient way to assess the reliability of clustering algorithms.
• This work highlights the potential of clustering algorithms for accurate biomolecular electrostatics computations.