W0016

Implementation of a Six-Dimensional Search for Difficult Molecular Replacement Problems. S. Sheriff, H. E. Klei, & M. E. Davis, Bristol-Myers Squibb Pharmaceutical Research Institute, P.O. Box 4000, Princeton, NJ 08543-4000

A six-dimensional molecular replacement procedure has been implemented using Perl scripts and the CCP4 version of the translation function of the molecular replacement program AMoRe. These tools have allowed the structure determination of (1) CTLA-4 monomer from NMR-derived coordinates; and (2) a potential complex of MurD with substrate. In both cases other molecular replacement programs, e.g. X-PLOR and AMoRe, when used in a conventional manner, either failed or did not yield an obvious solution. For the CTLA-4 monomer, the search was conducted on a fine grid of every 2.5 degrees yielding a total of 49 x 37 x 145 = 262,885 translation functions. The six-dimensional search took ~13 wall clock days in each of two enantiomorphic space groups. However, each search was split between three processors (Silicon Graphics R8000) so that the elapsed time was ~4.5 days. For MurD, two of the three domains were easily locatable by conventional molecular replacement, but the third domain required a six-dimensional search. In this case, a 5 degree search yielded a total of 19 x 19 x 73 = 26,335 translation functions. The six-dimensional search was conducted with the first two domains fixed in the translation function and the wall clock time was ~4 days, although the search was again divided over three processors so that the elapsed time was ~1.3 days. The processor used for these calculations, although fast, has been surpassed by other processors. Therefore, as processor speeds continue to improve, the six-dimensional search will become even more feasible. Moreover, the six-dimensional search is eminently parallelizable (albeit in its current state by hand) and therefore can be run over any accessible set of computers, which can reduce the elapsed time of the search considerably.