SEARCHING CONFORMATIONAL SPACE WITH THE SPECTRAL DISTANCE GEOMETRY ALGORITHM

The spectral distance geometry algorithm reported here contains a numb er of features that are not common to other distance geometry algorith ms used to determine molecular conformations from nuclear magnetic res onance data. A central question to ask for all algorithms in this area is whether the algorithm samples the conformation space well. In orde r to begin to answer this question we examine the sampling properties of the algorithm applied to unconstrained polypeptide chains. A new al gorithm to find better starting values is introduced. This starting al gorithm improves the sampling properties of our approach. From an init ial investigation of this improved algorithm it appears that its sampl ing properties are superior to those of similar methods that have been reported if no metrization is used. Other good qualities of the algor ithm are its speed and the fact that almost all starting values conver ge to an acceptable structure, which contrasts with a reported failure rate of approximately 20% for other distance geometry algorithms.