Statistical potentials extracted from protein structures: Are these meaningful potentials?

We present an inventory of factors that determine the stability of proteins. It is shown that the (true) potential of mean force between pairs of amino acid residues does not belong to this inventory. Therefore, such quantities are not needed for either estimating the stability of a protein or for predicting its structure. It is also shown that the so-called ''statistical potential'' as derived from the data bank of protein structures is neither the potential nor the potential of mean force for pairs of amino-acid residues. A critical examination of the nonadditivity of the many body potential of the mean force is also presented. (C) 1997 American Institute of Physics.