Standard Article

Score functions for structure prediction

Part 4. Bioinformatics

4.6. Methods for Structure Analysis and Prediction

Specialist Review

  1. Richard A. Goldstein

Published Online: 15 APR 2005

DOI: 10.1002/047001153X.g406202

Encyclopedia of Genetics, Genomics, Proteomics and Bioinformatics

Encyclopedia of Genetics, Genomics, Proteomics and Bioinformatics

How to Cite

Goldstein, R. A. 2005. Score functions for structure prediction. Encyclopedia of Genetics, Genomics, Proteomics and Bioinformatics. 4:4.6:67.

Author Information

  1. National Institute for Medical Research, London, UK

Publication History

  1. Published Online: 15 APR 2005


Protein structure prediction methods often make use of a score function, which describes how compatible a given fold is for the target sequence. There are a wide variety of methods for constructing these score functions, ranging from physics-based potentials that try to calculate the free energy, to simplified potentials, to functions drawn from machine-learning techniques. Often, these score functions have a number of parameters whose values must be determined. For physics-based potentials, these may be available through ab initio calculations and experimental observations of small molecules. For all of the different functional forms, it is possible to determine the values of these parameters on the basis of an analysis of available protein structures. Three different methods to do this – based on thermodynamics, statistics, and machine learning – are described.


  • tertiary structure prediction;
  • potentials of mean force;
  • protein folding;
  • Bayesian statistics;
  • molecular dynamics;
  • homology modeling;
  • fold recognition