UPGMA

UPGMA (Unweighted Pair-Group Method using Arithmetic Averages) is a simple bottom-up data clustering method used in bioinformatics for the creation of phylogenetic trees. The input data is a collection of objects with their pairwise distances and the output is a rooted tree (dendrogram). It is sometimes used for creating rooted phylogenetic trees under the assumption of a constant evolutionary rate. Initially, each object is in its own cluster. At each step, the nearest two clusters are combined into a higher-level cluster. The distance between any two clusters A and B is taken to be the average of all distances between pairs of objects a in A and b in B. UPGMA is not a well-regarded method for inferring phylogenetic trees unless the constant-rate assumption (molecular clock hypothesis) has been tested and justified for the data set being used.

UPGMA involves clustering of closely distant species. At each stage of clustering, tree branches are being built, and the branch lengths are calculated. UPGMA assumes a constant evolutionary rate, and so the two species in a cluster are given the same branch length from the node. It is a simple and fast method; however, because of the assumption, it often produces incorrect topologies when the assumption is not met.

Methods Using Distance Matrices

There are various methods of the distance matrix method. Listed below are the four main ones (Nei & Kumar, 2000):

UPGMA 

see above

Least Squares (LS) Method 

calculates the differences between the observed and estimated branch lengths between species. After it evaluates all possible topologies, it chooses the topology with the smallest difference. The estimation of branch lengths has two methods, Fitch-Margoliash and Least Squares.

Minimum Evolution (ME) Method 

estimates the total branch length of each topology. After it evaluates all possible topologies, it chooses the topology with the least total branch length. This method is computationally intensive and therefore slow, and with a small number of species to compare, the NJ method usually gives the same result as the ME method in less time.

Neighbour-Joining (NJ) Method 

involves clustering of neighbour species that are joined by one node. It does not evaluate all the possible tree topologies, but at each stage of clustering the ME method is used. Thus, the NJ method is considered a simplified version of the ME method.

References

• Nei M and Kumar S (2000). Molecular Evolution and Phylogenetics. Oxford University Press (New York; pp73-113).