A significant part of the problem of comprehending biological structure is the quantification of individual interactions. The large number of protein structures provides a rich set of data to use for understanding molecular interactions. These can inform us about the interactions in the biological milieu. We have several important new findings: (1) At close approach, <4.0 A, polar pairs of interacting residues dominate; (2) In an intermediate range, 4.0 A to 6.5 A, hydrophobic pairs dominate; (3) Amino acid side chains inside a protein pack in a highly regular way; whereas backbone elements pack less regularly; (4) Residue interactions derived from NMR structures differ significantly from those taken from X-ray structures; (5) Short range interactions along the backbone complement the long range interactions; (6) Amino acid-base interactions can be derived from limited structures and experimental binding data; (7) Intermolecular interactions can be quantified in similar ways and used in selecting drugs from databases, for modifying existing drugs, or a priori drug design. In the latter case, experimental screening of suggested new compounds is proving the approach to be more effective than previous approaches. Another part of comprehending biological macromolecular structures is the efficient and thorough sampling of the widest possible range of conformations. We have been pursuing this in several ways: (1) with new, much more efficient ways to generate conformations on lattices, (2) with protein threading schemes to place unknown sequences on known structures that include gaps and insertions, and (3) with methods to seek out proteins that bind to a given DNA or RNA sequence. The selection of viable conformations is made with the interactions derived above. By evaluating both the short range and long range interactions, tests on small proteins are showing that correct conformations can be selected with high precision.
