Neighbour-joining method

Fig. 2.1. A tree representing the phylogeny of Wolbachia in arthropods (groups A and B) and filarial nematodes (groups C and D). Group designations correspond to those proposed by Werren etal. (1995) and by Bandi etal. (1998). The names at the terminal nodes are those of the host species. The tree is based on the ftsZgene sequence alignment used by Bandi etal. (1998). The tree was obtained using a distance matrix method (Jukes and Cantor correction neighbour-joining method).

Saitou, N. and Nei, M. (1987) The neighbour-joining method a new method for reconstructing phylogenetic trees. Molecular biology and evolution, 4, 406 25. 

The basic principle of the join cost is to use the features from the two units to decide whether the join will be good or not. While this forms the basis of most calculations, it is common to employ an additional trick whereby we use a separate route and always assign a zero join cost to any two units which were neighbouring in the original data. As these two units were joined originally, we know that they form a perfect join and receive a cost of 0, regardless of what the features or join cost proper tells us. This is a powerful addition, and it biases the search to choose sequences of units that were originally concatenated. Thus while nearly all the joins we consider in the search are chosen by the join cost proper, when we consider the joins of the units which are actually chosen, we find many have be calculated by this other method. 

---