Statistical significance of protein structure prediction by threading

Area of Science:

• Computational biology
• Structural bioinformatics
• Statistical mechanics

Background:

• Protein structure prediction is crucial for understanding biological function.
• Threading methods align sequences to known structures, but assessing statistical significance remains challenging.
• Current methods often require computationally expensive simulations like sequence shuffling.

Purpose of the Study:

• To develop a universal statistical measure for evaluating protein structure prediction accuracy in threading.
• To introduce a single parameter, epsilon, for quantifying the probability of a native-like alignment.
• To provide a method for optimizing threading simulations and gap penalties.

Main Methods:

• Introduction of a single parameter, epsilon, as a universal measure of statistical significance.
• Epsilon calculation based on query sequence length, composition, and decoy number.
• Theoretical analysis and comparison with gapless threading results.
• Estimation of decoy numbers required for native structure identification using interaction potentials.

Main Results:

• Epsilon provides a direct measure of alignment significance, independent of sequence shuffling.
• The study estimates the number of decoys needed to reliably identify native structures with current potentials.
• The framework allows for the extension to determine optimal gap penalties for various threading methods.

Conclusions:

• The epsilon parameter offers an efficient and accurate way to assess statistical significance in protein structure prediction by threading.
• This work provides a foundation for optimizing threading algorithms and improving their performance.
• The developed method reduces computational cost while enhancing the reliability of structure prediction.