WEIGHTING IN SEQUENCE SPACE - A COMPARISON OF METHODS IN TERMS OF GENERALIZED SEQUENCES

Four methods for weighting aligned biological sequences have recently appeared that differ mathematically, philosophically, and in their res ults. Thus, while there is consensus about the need to weight sequence s, the method to use is contentious. A geometric analysis based on a c ontinuous sequence space is presented that provides a common framework in which to compare the methods. It is concluded that there are two ' 'best'' methods. When the sequences are known to be phylogenetically r elated and a tree can be generated without introducing excessive stres s into the data, the method of Altschul et al. [Altschul, S. F., Carro ll, R. J. & Lipman, D. J. (1989) J. Mol. Biol. 207, 647-653] is approp riate. When the sequences are not known to be phylogenetically related or a tree cannot be produced without unduly distorting the distances between the sequences, a modification of the method of Sibbald and Arg os [Sibbald, P. R. & Argos, P. (1990) J. Mol. Biol. 216, 813-818] is p referable.