W0311

Fast Molecular Replacement by Maximum Likelihood in Phaser. Airlie J. McCoy¹, Laurent C. Storoni¹, Ralf Grosse-Kunstleve² & Randy J. Read¹, ¹Cambridge Inst. for Medical Research, Wellcome Trust/MRC Building, Hills Rd., Cambridge, CB2 2XY, U.K., ²Lawrence Berkeley Laboratory, 1 Cyclotron Rd., MS 4-230, Berkeley, CA 94720, USA.

Experiences with the molecular replacement program Beast (1) have shown that maximum likelihood rotation and translation targets are more accurate and more sensitive to the correct orientation and position than traditional targets. But this comes at a computational cost: brute-force rotation and translation searches can take hours to days to run.

Following the spirit of a suggestion by Bricogne (2), we have used series approximations of the likelihood functions to derive fast likelihood-based molecular replacement targets that can be computed by FFT. These have been implemented in a new program, Phaser, along with the full likelihood-based targets. With this in place, we can use the strategy Navaza devised for AMoRe (3): a fast score is used to find plausible solutions, which are then rescored by a more sensitive but more expensive method, in this case likelihood. Phaser now provides the benefits of Beast, but runs in minutes.

Experiences with difficult molecular replacement test cases will be used to illustrate the advantages of Phaser. Apart from increased sensitivity, Phaser also makes good use of the information from partial solutions of the molecular replacement problem.

2. Bricogne, G. (1992). Proceedings of the CCP4 Study Weekend on Molecular Replacement (eds. W. Wolf, E.J.Dodson & S. Gower), pp. 62-75. Warrington: Daresbury Laboratory.