research papers 42 Knight  RSPS Acta Cryst. (2000). D56, 42±47 Acta Crystallographica Section D Biological Crystallography ISSN 0907-4449 RSPS version 4.0: a semi-interactive vector-search program for solving heavy-atom derivatives Stefan D. Knight Uppsala Biomedical Center, Department of Molecular Biology, Box 590, S-751 24 Uppsala, Sweden Correspondence e-mail: stefan@xray.bmc.uu.se # 2000 International Union of Crystallography Printed in Denmark ± all rights reserved A program for inspection and interpretation of the Patterson function is described. The program is mainly intended for ®nding heavy-atom positions from difference Patterson maps, but may also be used to locate molecules with non-crystal- lographic symmetry when the local axis is nearly parallel to a crystallographic symmetry axis. Options are available for vector-based methods to locate heavy-atom sites, for ®nding sets from a list of possible heavy-atom positions and for checking of potential solutions. Both crystallographic and non-crystallographic symmetry may be used, either indepen- dently or in conjunction. Received 20 April 1999 Accepted 11 October 1999 1. Introduction Finding heavy-atom positions in at least one heavy-atom derivative is a crucial step in any de novo protein structure determination using isomorphous replacement and/or anom- alous scattering methods. Traditionally, heavy-atom deriva- tives have been solved by inspection of the difference Patterson function based on the derivative and native data, and in many cases this is still an ef®cient method of locating one or a few heavy-atom sites. However, with increasing complexity, following from larger structures and/or high symmetry in the crystal lattice, an automated system for difference Patterson interpretation becomes more and more indispensable. In the case of high symmetry, be it crystal- lographic (space-group symmetry; SGS) or non-crystal- lographic symmetry (NCS), the increased complexity arises simply because of the crowding of numerous peaks in the Patterson function. Larger structures give rise to a higher background owing to an increased number of protein±protein and protein±heavy-atom vectors, thus obscuring the already weak signal from the heavy atom(s). Furthermore, with bigger structures the likelihood of multiple heavy-atom binding sites increases. This in turn has two effects. Firstly, the number of peaks in the difference Patterson function increases as N 2  N (where N is the number of sites in the unit cell), so that crowding of peaks and possible peak overlap becomes more and more serious. Secondly, each additional site adds to the background of the Patterson function, whereas the signal per site remains constant. Automated methods for solving difference Patterson func- tions may be loosely divided into two categories: vector-search methods (for some examples, see Tickle, 1983; Terwilliger et al., 1987; Knight, 1989; Knight et al., 1990) and superposition methods (Buerger, 1959; Sheldrick, 1991). RSPS is based on vector-search methods, which are essentially trial-and-error approaches. The Patterson function is sampled at points corresponding to the vectors predicted from trial solutions,