Polarity as a criterion in protein design

Hypothetical proteins can be tested computationally by determining whether or not the designed sequence-structure pair has the characteristics of a typical globular protein. We have developed such a test by deriving quantities with approximately constant value for all globular proteins, based on empirical analysis of the exposed and buried surfaces of 128 structurally known proteins. The characteristic quantities that best appear to segregate badly designed or deliberately misfolded proteins from their properly folded natural relatives are the polar fraction of side chains on the protein surface and, independently, in the protein interior. Three of the seven hypothetical structures tested here can be rejected as having too many polar side-chain groups in the interior or too few on the protein surface. In addition, a recently designed nutritional protein is identified as being very much unlike globular proteins. These database-derived characteristic quantities are useful in screening designed proteins prior to experiment and may be useful in screening experimentally determined (X-ray, NMR) protein structures for possible errors.