An extended dynamic programming algorithm is presented that is applica
ble to the fragment assembly phase of the site mapping fragment assemb
ly approach to peptide docking. After constructing a free energy map o
f the receptor using each of the amino acids in the peptides to be doc
ked, we apply the algorithm to two systems: HIV-1 protease complexed w
ith a synthetic hexameric inhibitor, and MHC HLA-A2 complexed with a n
onameric peptide. The all atom root mean square deviation between the
predicted and crystal structures was 1.7 and 2.0 Angstrom, respectivel
y. While these results are reasonable considering the relatively coars
e level of mapping, the more important result is that the structures a
re probably very close to the best obtainable by an exhaustive search
through the entire data map, and yet are obtained with a reduction of
3-5 orders of magnitude in the number of computations. We also outline
a prescription for an iterative procedure which finds the global mini
mum with increasing confidence. (C) 1996 by John Wiley & Sons, Inc.