User guide

Quantity One User Guide
19-14
Neighbor Joining
This type of phylogenetic tree is computed is based on minimizing the total
branch length at each stage of clustering. The method also finds branch
lengths between nodes. The approximate distance between any two samples
in this tree can be found by adding the branch lengths that connect the
samples.
Other Methods
1
,
2
The following methods are based on the algorithm below:
1. Begin with n clusters—one cluster for each sample.
2. Compute the similarity matrix for the samples.
3. Convert the similarity matrix into a distance matrix d using the
appropriate distance formula.
4. Join the two clusters with the minimum distance into one cluster.
Compute the similarity value for this cluster.
5. Recompute the distance matrix d using the cluster that was formed in
Step 4.
Steps 4 and 5 are repeated until there is only one cluster. The difference
between the methods below is based on the definition of minimum distance
in Step 4, and on the method of computing the new distance matrix in step 5.
In the discussion that follows, let:
p, q be indices indicating two clusters that are to be joined into a single
cluster.
k be the index of the cluster formed by joining clusters p and q.
i be the index of any remaining clusters other than cluster p, q, or k.
1.For a complete explanation of the calculations and assumptions used to
generate these dendrograms, please refer to Sneath and Sokal. Numerical
Taxonomy, San Francisco: W. H. Freeman & Company, 1973.
2. Vogt and Nagel, Clinical Chemistry 38 (2): 182-198, (1992).