Effective interactions cannot replace solvent effects in a lattice model of proteins.

Protein folding and protein design are among the most challenging problems of the past ten years in biophysics and molecular biology. For a given protein, it is possible to extract, from existing protein databases, a set of specific (i.e., belonging to the investigated protein) effective amino-acid (AA) interactions able to stabilize the native state. On the other hand, attempts to find global effective AA interactions, which would be able to stabilize all proteins at once, failed. Using a simple lattice model where the solvent degrees of freedom are (semi)explicitly taken into account, we show that the absence of global effective AA interactions is due to the solvent and that on this lattice model the solvent effects cannot be reproduced by amino-acid effective interactions.